Download Ch4 Pre-release version - Arctic Climate Impact Assessment

Chapter 4
Future Climate Change: Modeling and Scenarios for the Arctic
Lead Authors
Vladimir M. Kattsov, Erland Källén
Contributing Authors
Howard Cattle, Jens Christensen, Helge Drange, Inger Hanssen-Bauer,Tómas Jóhannesen, Igor Karol, Jouni Räisänen, Gunilla Svensson,
Stanislav Vavulin
Consulting Authors
Deliang Chen, Igor Polyakov, Annette Rinke
Contents
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .100
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101
4.2. Global coupled atmosphere-ocean general circulation
models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .103
4.2.1. Equilibrium and transient response experiments . . . . . . . . . . . .104
4.2.2. Initialization and coupling issues . . . . . . . . . . . . . . . . . . . . . . . . . .104
4.2.3. Atmospheric components of AOGCMs . . . . . . . . . . . . . . . . . . . .105
4.2.4. Ocean components of AOGCMs . . . . . . . . . . . . . . . . . . . . . . . . .106
4.2.5. Land-surface components of AOGCMs . . . . . . . . . . . . . . . . . . . .107
4.2.6. Cryospheric components of AOGCMs . . . . . . . . . . . . . . . . . . . .107
4.2.7. AOGCMs selected for the ACIA . . . . . . . . . . . . . . . . . . . . . . . . .108
4.2.8. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .109
4.3. Simulation of observed arctic climate with the ACIAdesignated models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .109
4.3.1. Observational data and reanalyses for model evaluation . . . . .110
4.3.2. Specifying the ACIA climatological baseline . . . . . . . . . . . . . . . . .112
4.3.3. Surface air temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .113
4.3.4. Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .114
4.3.5. Other climatic variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .115
4.3.6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .117
4.4. Arctic climate change scenarios for the 21st century
projected by the ACIA-designated models . . . . . . . . . . . . . .118
4.4.1. Emissions scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .120
4.4.2. Changes in surface air temperature . . . . . . . . . . . . . . . . . . . . . . .121
4.4.3. Changes in precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .126
4.4.4. Changes in other variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .127
4.4.5. ACIA-designated models in the CMIP2 exercise . . . . . . . . . . . . .128
4.4.6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .129
4.5. Regional modeling of the Arctic . . . . . . . . . . . . . . . . . . . . . . .129
4.5.1. Regional climate models of the arctic atmosphere . . . . . . . . . . .130
4.5.1.1. General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .130
4.5.1.2. Simulations of present-day climate with regional
climate models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .131
4.5.1.3.Time-slice projections from atmospheric RCMs . . . . . . .132
4.5.2. Regional Arctic Ocean models . . . . . . . . . . . . . . . . . . . . . . . . . . .134
4.5.3. Coupled arctic regional climate models . . . . . . . . . . . . . . . . . . .135
4.5.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .135
4.6. Statistical downscaling approach and downscaling of
AOGCM climate change projections . . . . . . . . . . . . . . . . . . .136
4.6.1. Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136
4.6.1.1. Predictands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136
4.6.1.2. Predictors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136
4.6.1.3. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136
4.6.1.4. Comparison of statistical downscaling and regional
modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .136
4.6.2. Statistical downscaling of AOGCM climate change projections
in the Arctic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .137
4.6.3. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .140
4.7. Outlook for improving climate change projections for the
Arctic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .140
4.7.1.The Arctic part of the climate system – a key focus in
developing AOGCMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .140
4.7.2. Improved resolution of arctic processes . . . . . . . . . . . . . . . . . . .141
4.7.3. Better representation of the stratosphere in AGCMs . . . . . . . .142
4.7.4. Coupling chemical components to GCMs . . . . . . . . . . . . . . . . .142
4.7.5. Ensemble simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .143
4.7.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .143
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .144
Pre-release Version 24 March, 2005
www.acia.uaf.edu
100
Arctic Climate Impact Assessment
Summary
Increased atmospheric concentrations of greenhouse
gases (GHGs) are very likely to have a larger effect on
climate in the Arctic than anywhere else on the globe.
Physically based, global coupled atmosphere-land-ocean
climate models are used to project possible future climate change. Given a change in GHG concentrations,
the resulting changes in temperature, precipitation, seasonality, etc. can be projected. Future emissions of
GHGs and aerosols can be estimated by making assumptions about future demographic, socioeconomic, and
technological changes.The Intergovernmental Panel on
Climate Change (IPCC) prepared a set of emissions scenarios for use in projecting future climate change.This
assessment uses the A2 and B2 scenarios, which are in
the middle of the range of scenarios provided by the
IPCC. Projections from the IPCC climate models indicate a global mean temperature increase of 1.4 ºC by the
mid-21st century compared to the present climate for
both the A2 and B2 scenarios (IPCC, 2001).Toward the
end of the century, the global mean temperature
increase is projected to be 3.5 ºC and 2.5 ºC for the two
scenarios, respectively.
Over the Arctic, the ACIA-designated models project a
larger mean temperature increase: for the region north
of 60º N, both scenarios result in a 2.5 ºC increase by
the mid-21st century. By the end of the 21st century,
arctic temperature increases are projected to be 7 ºC
and 5 ºC for the A2 and B2 scenarios, respectively, compared to the present climate. By then, in the B2 scenario, the models project temperature increases of
around 3 ºC for Scandinavia and East Greenland, about 2
ºC for Iceland, and up to 5 ºC for the Canadian
Archipelago and Russian Arctic.The five-model mean
warming over the central Arctic Ocean is greatest in
autumn and winter (up to 9 ºC by the late 21st century
in the B2 scenario), as the air temperature reacts strongly to reduced ice cover and thickness. Average autumn
and winter temperatures are projected to rise by 3 to 5
ºC over most arctic land areas by the end of the 21st
century. By contrast, summer temperature increases
over the Arctic Ocean are projected to remain below 1
ºC throughout the 21st century.The contrast between
greater projected warming in autumn and winter and
lesser warming in summer also extends to the surrounding land areas but is less pronounced there. In summer,
the projected warming over northern Eurasia and northern North America is greater than that over the Arctic
Ocean, while in winter the reverse is projected. All of
the models suggest substantially smaller temperature
increases over the northern North Atlantic sector than in
the other parts of the Arctic.
By the late 21st century, projected precipitation increases in the Arctic range from about 5 to 10% in the
Atlantic sector to as much as 35% in certain high Arctic
locations (for the B2 scenario). As for temperature, the
projected increase in precipitation is generally greatest
in autumn and winter and smallest in summer.
A slight decrease in pressure in the polar region is projected for throughout the year.While impact studies
would benefit from projections of wind characteristics
and storm tracks in the Arctic, available analyses in the
literature are insufficient to justify firm conclusions
about possible changes in the 21st century.
The models also project a substantial decrease in snow
and sea-ice cover over most of the Arctic by the end of
the 21st century.
The projected increase in arctic temperatures is accompanied by large between-model differences and considerable interdecadal variability. Dividing the average projected temperature change by the magnitude of projected variability suggests that, despite the large warming
projected for the Arctic, the signal-to-noise ratio is actually lower in the Arctic than in many other areas.
The Arctic is a region characterized by complex and
insufficiently understood climate processes and feedbacks, contributing to the challenge that the Arctic poses
from the view of climate modeling. Several weaknesses
of the models related to descriptions of high-latitude
surface processes have been identified, and these are
among the most serious shortcomings of present-day
arctic climate modeling.
Local and regional climate features, such as enhanced precipitation close to steep mountains, are not well represented in global climate models due to the limited horizontal resolution of the models.To describe local climate,
physical modeling or statistically based empirical links
between the large-scale flow and local climate can be
used. Despite rapid developments in arctic regional climate modeling, the current status of developments in this
field did not allow regional models to be used as principal
tools for the ACIA.Therefore, the ACIA used projections
from coupled global models, either directly or in combination with statistical downscaling techniques.
A model simulation provides one possible climate scenario.This is not a prediction of future climate change, but
a projection based on a prescribed change in the concentration of atmospheric GHGs. A climate shift can be
caused by natural variability as well as by changes in GHG
concentrations. Natural variability in the Arctic is large and
could mask or amplify a change resulting from increased
atmospheric GHG concentrations.To assess the relative
importance of natural variability versus a prescribed climate forcing, an ensemble of differently formulated climate models should be used. For this assessment, five different models are used to give an indication of simulation
uncertainty versus forced changes, although greater numbers of simulations would provide a better estimate of climate change probability distributions, and perhaps allow
the estimation of changes in the frequency of winter
storms, and temperature and precipitation extremes, etc.
While the level of uncertainty in climate simulations can
probably be reduced with improved model formulations,
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
it will never be certain that all physical processes relevant
to climate change have been included in a model simulation.There can still be surprises in the understanding of
climate change.The projections presented here are based
on the best knowledge available today about climate
change; as climate-change science progresses there will
always be new results that may change the understanding
of how the arctic climate system works.
4.1. Introduction
To assess climate change impacts on societies, ecosystems, and infrastructure, possible changes in physical climate parameters must first be projected.The physical climate change projections must in turn be calculated from
changes in external factors that can affect the physical climate. Examples of such factors include atmospheric composition, particularly atmospheric concentrations of
GHGs and aerosols, and land-surface changes (e.g.,
deforestation).This chapter describes the options available to make such projections and their application to the
Arctic.The main emphasis is on physically based models
of the climate system and the relationship between global
climate change and regional effects in the Arctic.
Physically based climate models are used to obtain climate scenarios – plausible representations of future climate that are consistent with assumptions about future
emissions of GHGs and other pollutants (i.e., emissions
scenarios) and with present understanding of the effects
of increased atmospheric concentrations of these components on the climate (IPCC-TGCIA, 1999).
Correspondingly, by using a climate change scenario, the
difference between the projected future climate and the
current climate is described. Being dependent on sets of
prior assumptions about future human activities, demographic and technological change, and their impact on
atmospheric composition, climate change scenarios are
not predictions, but rather plausible, internally consistent descriptions of possible future climates.
In addition to physical climate modeling, there are alternative methods for providing climate scenarios for use in
impact assessments.These include synthetic scenarios (also
referred to as arbitrary or incremental scenarios) and analogue scenarios. None of the alternatives provide a physically consistent climate change scenario including both
atmospheric composition changes and physically coupled
changes in temperature, precipitation, and other climate
variables. Nevertheless, due to their relative simplicity
they can be useful and adequate for some types of impact
studies.There are also climate scenarios that do not fall
into any of these categories, which primarily employ
extrapolation of either ongoing trends in climate, or
future regional climate, on the basis of projected global or
hemispheric mean climate change. A separate group of
scenarios is based on expert judgments. All of the methods have their limitations, but each has some particular
advantages (see Mearns et al., 2001; Carter et al., 2001).
101
Synthetic scenarios are based on incremental changes in
climatic variables, particularly air temperature (e.g., +1,
+2, +3 ºC) and precipitation (e.g., +5, +10, +15%). Such
scenarios often assume a uniform annual change in the
variables over the region under consideration; however,
some temporal and spatial variability may be introduced as
well. Synthetic scenarios provide a framework for conducting sensitivity studies of potential impacts of climate
change using impact models. Careful selection of the
range and combinations of changes (e.g., using knowledge
based on climate model projections), can facilitate “guided” sensitivity analysis, enabling an examination of both
the modeled behavior of a system under a plausible range
of climatic conditions and the robustness of impact models applied under changed and often unprecedented environmental conditions. Synthetic scenarios can provide a
useful context for understanding and evaluating responses
to more complex scenarios based on climate model outputs.Transparency to users and limited computational
resource requirements, which allow examination of a
wide range of potential climate changes (the range is further increased by the possibility of changing individual
variables independent of one another), are among the
advantages of synthetic scenarios.Their main disadvantage
is the lack of internal consistency in applying uniform
changes over large and highly variable areas such as the
Arctic. Arbitrary changes in different variables may also
lead to inconsistencies in synthetic scenarios that can limit
their applicability and appropriateness. In addition, synthetic scenarios are not directly related to GHG forcing.
Analogue scenarios of a future climate are of two types:
temporal analogue scenarios, which are based on previous warm climate conditions (determined either by
instrumental or proxy data), and spatial analogue scenarios, which are based on current climate conditions in
warmer regions.The use of historic instrumental
records is an apparent advantage of the past climate analogues over other approaches. However, the availability
of historic observational data for the Arctic is extremely
limited. Proxy climate data, while representing in some
cases a physically plausible climate different from the
current climate to a degree similar to that of the climate
projected for the 21st century, are also not available for
many locations.The quality of geological records is often
uncertain, and the resolution coarse. Furthermore, the
paleoclimate changes are unlikely to have been driven by
an increase in GHG concentrations. Spatial analogues are
also unrelated to GHG forcing and are often physically
implausible.The lack of availability of proper analogues
is the major problem for the analogue scenario
approach.The IPCC recommends that analogue scenarios are not used, at least not independently of other
types of scenario (Carter et al., 1994).
Physical climate models are based on the laws of physics
and discrete numerical representations of these laws that
allow computer simulations.Trenberth (1992) describes
how climate models can be constructed and their underlying physical principles. Of the hierarchy of climate
models (Box 4.1), global coupled atmosphere-ocean
Pre-release Version 24 March, 2005
www.acia.uaf.edu
102
Arctic Climate Impact Assessment
general circulation models (AOGCMs) are widely
acknowledged as the principal, and most promising rapidly developing tools for simulating the response of the
global climate system to increasing GHG concentrations.
In its Third Assessment Report, the IPCC (2001) concluded that state-of-the-art AOGCMs in existence at the
turn of the century provided “credible simulations of climate, at least down to subcontinental scales and over
temporal scales from seasonal to decadal”, and as a class
were “suitable tools to provide useful projections of the
future climate” (McAvaney et al., 2001).The IPCC
(2001) identified the following primary sources of
uncertainty in climate scenarios based on AOGCM projections: uncertainties in future emissions of GHGs and
aerosols (emissions scenarios), and in conversion of the
emissions to atmospheric concentrations and to radiative
forcing of the climate; uncertainties in the global and
regional climate responses to emissions simulated by different AOGCMs; and uncertainties due to inaccurate
representation of regional and local climate. A disadvantage of the AOGCMs as a tool for constructing scenarios
is their high demand for computational resources, which
makes it expensive and time-consuming to carry out calculations for multiple emissions scenarios.
The selection of climate scenarios for impact assessments is always controversial and vulnerable to criticism
(Smith et al., 1998). Mearns et al. (2001) suggest that,
to be useful for impact assessments and policy makers,
climate scenarios should be consistent with global projections at the regional level (i.e., projected changes in
regional climate may lie outside the range of global
mean changes but should be consistent with theory and
model-based results); be physically plausible and realistic; provide a sufficient number of variables and appropriate temporal and spatial scales for impact assessments; be representative, reflecting the potential range
of future regional climate change; and be accessible.
Compared to the other methods of constructing climate
change scenarios, only AOGCMs (possibly in combination with dynamic or statistical downscaling methods)
have the potential to provide spatially and physically consistent estimates of regional climate change due to
increased atmospheric GHG concentrations (IPCCTGCIA, 1999).The AOGCM projections are available
for a large number of climate variables, at a variety of
temporal scales, and for regular grid points all over the
world, which should be sufficient for many impact
Box 4.1. Climate model hierarchy
Climate models have very different levels of complexity with respect to resolution and comprehensiveness.
Available computing resources may limit model complexity for practical reasons, but the scientific question to be
addressed is the main factor determining the required model complexity. Different levels of reduction (or simplification) create a hierarchy of climate models (McAvaney et al., 2001).
Simple climate models of the energy-balance type, with zero (globally averaged) to two (latitude and height) spatial dimensions, belong to the lowest level of the hierarchy. Based upon parameters derived from more complex
climate models, they are useful in studies of climate sensitivity to a particular process over a wide range of parameters (e.g., in a preliminary analysis of climate sensitivity to various emissions scenarios, see section 4.4.1). Simple
climate models can also be used as components of integrated assessment models, for example, in analyses of the
potential costs of emission reductions or impacts of climate change (see Mearns et al., 2001).
Earth system models of intermediate complexity (EMICs) bridge the gap between the simple models and the
comprehensive three-dimensional climate models (see Claussen et al., 2002).These models explicitly simulate
interactions between different components of the climate system; however, at least some of the components have
a reduced complexity, potentially limiting their applicability.These models are computationally efficient, allow for
long-term climate simulations measured in thousands and tens of thousands of years, and are primarily used for
studies of particular climate processes and feedbacks that are not believed to be affected by the dynamical simplifications introduced.
Comprehensive three-dimensional coupled atmosphere-ocean general circulation models (AOGCMs) occupy the
top level of the hierarchy.The term “general circulation” refers to large-scale flow systems in the atmosphere and
oceans, and the associated redistribution of mass and energy in the climate system. General circulation models
(GCMs) simulate the behavior of these systems and the interactions between them and with other components
of the climate system, such as sea ice, the land surface, and the biosphere. Atmosphere-ocean general circulation
models are widely acknowledged as the most sophisticated tool available for global climate simulations, and particularly for projecting future climate states.
Atmosphere-ocean general circulation models were preceded by far less computationally demanding atmospheric
GCMs coupled to simple parameterizations of the upper mixed layer of the ocean (AGCM/OUML), which still
play an important role in studies of processes and feedbacks in the climate system (see also section 4.2.1) and in
paleoclimate simulations.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
103
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
assessments. Employing an ensemble of different models
increases the representativeness of AOGCM-based scenarios.When AOGCMs are used to provide the central
scenarios, they can be combined with other types of scenarios (e.g., with synthetic scenarios applied at the
regional level, for which the AOGCMs provide a physically plausible range of climate changes).
For this assessment, five AOGCMs (referred to as the
ACIA-designated models, see section 4.2.7) were selected for constructing future climate change scenarios for
the Arctic (see section 1.3.2).The ACIA-designated models are drawn from the generation of climate models evaluated by the IPCC (2001).This chapter begins with a
brief description of the state-of-the-art in AOGCM
development at the time of the IPCC assessment (section
4.2), followed by an evaluation of the ACIA-designated
models’ performance in simulating the current climate of
the Arctic (section 4.3). Projections of future climate
change in the Arctic using the ACIA-designated models
are the central focus of this chapter (section 4.4). An
assessment of possible climate change at scales smaller
than subcontinental, such as the scale considered by the
ACIA, requires the application of a downscaling tech-
nique to the AOGCM output (see Box 4.1). In this
assessment, two methods of downscaling AOGCM projections have been considered: regional climate modeling
(section 4.5), and statistical downscaling (section 4.6).
Finally, section 4.7 presents the outlook for improving
AOGCM-based climate change projections for the Arctic.
4.2. Global coupled atmosphere-ocean
general circulation models
The atmosphere, oceans, land surface, cryosphere, and
associated biology and chemistry form interactively coupled components of the total climate system. Climate
models are primary tools for the study of climate, its sensitivity to external and internal forcing factors, and the
mechanisms of climate variability and change.These
models attempt to take into account the various processes important for climate in the atmosphere, the oceans,
the land surface, and the cryosphere, as well as the interactions between them (Fig. 4.1). In addition, models are
increasingly incorporating components that describe the
role of the biosphere and chemistry in order to provide a
comprehensive description of the total earth system.
While the resolution of AOGCMs used for projections of future climate is rapidly improving, it is still insufficient to
capture the fine-scale structure of climatic variables in many regions of the world that is necessary for impact
assessment studies (Giorgi et al., 2001; Mearns et al., 2001). Hence, a number of techniques exist to enhance the
resolution of AOGCM outputs.These techniques fall into three categories:
• High- or variable-resolution stand-alone AGCM
simulations initialized using atmospheric and
land-surface conditions interpolated from the
corresponding AOGCM fields and driven with
the sea-surface temperature and sea-ice distributions projected by the AOGCM.
• High-resolution regional (or limited-area) climate models (RCMs) restricted to a domain
with simple lateral boundaries, at which they
are driven by outputs from GCMs or largerscale RCMs.
• Statistical downscaling methods that are based
on empirically derived relations between
observed large-scale climate variables and local
variables, and which apply these relations to the
large-scale variables simulated by GCMs (or
RCMs).
Each of the regionalization techniques is characterized
by its own set of advantages and disadvantages.
Giorgi et al. (2001) provide details on the high- and
variable-resolution AGCMs, while RCMs and statistical
downscaling are discussed in sections 4.5 and 4.6.
Climate model hierarchy and downscaling techniques
(based on Kattsov and Meleshko, 2004).
Pre-release Version 24 March, 2005
www.acia.uaf.edu
104
Arctic Climate Impact Assessment
Because physical processes and feedbacks play a key role
in the arctic climate system, this section focuses primarily
on the physical components of climate models. However,
land-surface biology is an important factor in determining the key thermal and radiative properties of the land
surface, surface hydrology, turbulent heat and gas
exchanges, and other processes. Likewise, the interaction
of ocean biology with physical processes is important for
air-sea gas exchange, including key processes related to
cloud formation such as dimethyl sulfide exchange.
Coupled AOGCMs are made up of component models of
the atmosphere, ocean, cryosphere, and land surface that
are interactively coupled via exchange of data across the
interfaces between them. For example, the ocean component is driven by the atmospheric fluxes of heat, momentum, and freshwater simulated by the atmospheric component.These heat and freshwater fluxes are themselves
functions of the sea-surface temperatures simulated by
the ocean model. Other driving fluxes for the ocean are
produced by the brine rejection that occurs during seaice formation, freshwater from sea-ice melt, and freshwater river discharge at the continental boundaries.
Atmosphere-ocean general circulation models are continually evolving.The state-of-the-art climate modeling
described in this section refers to the generation of
models from the late 1990s and very early 2000s, and is
close to that evaluated by the IPCC (2001).
4.2.1. Equilibrium and transient response
experiments
Early climate simulations were conducted using atmospheric models coupled to highly simplified representations of the ocean. In these models, only the upper
ocean was normally represented and then only as a simple fixed-depth slab of water some tens of meters deep
in which the temperature responded directly to changes
in atmospheric heat fluxes. Such models are still useful
for short sensitivity experiments, such as exploring the
impact of new representations of physical processes on
climate change in experiments in which the concentration of atmospheric GHGs is instantaneously doubled in
the model atmosphere.These models enable a quick
assessment of the “equilibrium response” of climate to a
given perturbation.The equilibrium response is the
change in climate resulting from a perturbation (e.g., a
specified increase in effective carbon dioxide (CO2) concentration) after a period long enough for the climate to
reach an equilibrium state. However, such models
assume that vertical and horizontal heat transports in the
ocean do not change when the climate changes.
Many centers have developed models with full dynamic
deep-ocean components over the past decade.The
dynamic oceans introduce the long timescales (multicentury to millennia) associated with the equilibration of
the abyssal ocean. Such long timescales are absent in the
models that represent the ocean as a shallow slab of
water. In particular, this development has enabled the
Fig. 4.1. Schematic illustrating the representation of the earth
system by a coupled atmosphere-ocean general circulation model.
Actual grid size and number of levels may vary.
exploration of the “transient response” of climate to
changing concentrations of GHGs, as well as the examination of many aspects of natural climate variability.The
transient response is the change over time as the perturbation (e.g., a continuous change in GHG concentrations) is applied. In the case of GHG-induced temperature change, the transient response is smaller than the
equilibrium response because the large thermal inertia
of the oceans slows the rate of warming.
4.2.2. Initialization and coupling issues
Owing to embodied feedbacks between ocean and atmosphere, an AOGCM-simulated climate is less constrained
than climates simulated by stand-alone atmospheric or
oceanic general circulation models (GCMs). Upon coupling, an AOGCM-simulated climate typically undergoes
a so-called coupling shock (fast drift due to imbalances
in the initial conditions between the component models
at the time of coupling) and then, after a close-tobalance state between the interacting components of the
AOGCM has been achieved, a gradual drift toward the
model’s equilibrium climatic state.The presence of climate drift, if it is significant, can complicate the study of
a possible climate change signal. For example, large
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
drifts can potentially distort the behavior of various
feedback processes present in the climate system and,
dependent on the mean state of the model, distort the
calculated climatic response to a given change in forcing.
The climate drift problem can introduce many technical
considerations into the application of AOGCMs.To limit
the influence of climate drift (especially the fast-drift
component), careful initialization of AOGCMs is very
important.This has led to a relatively wide array of initialization methods (e.g., Stouffer and Dixon, 1998).
Initialization techniques often include a sequence of runs
of component models separately, and in the coupled
mode the components are constrained by observations at
their interfaces.This makes it possible to reduce the climate drift and, particularly, the coupling shock.
Until recently, it has been necessary to use so-called
“flux adjustments” (or “flux corrections”, Sausen et al.,
1987) to prevent drift in the climate of the coupled system that arises from inadequacies in the component
models and in the simulated fluxes at their interfaces.
These adjustments are normally derived as fields of spatially varying “corrections” to the heat and freshwater
fluxes between the atmosphere and ocean components
of the model.They are often derived during a calibration
run of the AOGCM in which the sea-surface temperatures and surface salinities are constrained to observed
climatological values of these quantities.The flux adjustments are then applied to succeeding runs of the model
to provide improved simulation of the coupled system.
In some cases, flux adjustments have also been applied to
momentum fluxes.While flux adjustments have not been
applied over land, it has in the past been necessary to
flux-adjust the fields of sea-ice concentration and thickness. A driver in the development of coupled models has
been to improve models to the stage where they can run
without flux adjustments, as is now the case for some
AOGCMs. In the AOGCMs that continue to use this
technique, flux adjustments have become smaller as
models have improved. Interestingly, the IPCC did not
find systematic differences in the simulation of internal
climate variability between flux-adjusted and non-flux
adjusted AOGCMs (McAvaney et al., 2001), thus supporting the use of both types of model in the detection
and attribution of climate change.
4.2.3. Atmospheric components of AOGCMs
The atmospheric component of AOGCMs enables simulation of the evolution with time of the spatial distributions of the vector wind, temperature, humidity, and surface pressure.This is done by discretization of the basic
equations governing the behavior of the atmosphere, and
implementation of these discretized equations on an
appropriate computer.The equations are time-stepped
forward at intervals that typically vary from a few minutes to tens of minutes, depending on the model formulation and resolution, to produce an evolving simulation
of the behavior of the atmospheric flow and associated
temperature, humidity, and surface-pressure fields.
105
The model dynamics are usually represented either as
periodic functions defined as the sum of several waves
(spectral models) or on a grid of points (finite-difference
models) covering the globe for various levels of the
atmosphere.Typically, the atmospheric components of
the generation of climate models evaluated by the IPCC
(2001) operated on grids with a horizontal spacing of
200 to 300 km and 10 to 20 vertical levels.Various
schemes are available for the specification of vertical
coordinates (e.g., Kalnay, 2003).
Simulation of some climatic variables in high latitudes
(e.g., atmospheric moisture) using global models presents
certain problems. Finite-difference GCMs require undesirable filtering operations in order to avoid computational instability when a reasonable time step is used in
regions of converging meridians such as the Arctic.While
polar filtering is not needed in spectral models, these
models produce fictitious negative moisture amounts in
the dry high-latitude atmosphere, thus calling for correction procedures. Both problems are apparently overcome
by the application of semi-Lagrangian schemes for moisture advection, which are used in a number of atmospheric general circulation models (AGCMs). However,
in semi-Lagrangian schemes, the advantages of large time
steps and the absence of spurious negative moisture values are partially offset by the lack of exact moisture conservation. New schemes have recently started to appear
that combine the semi-Lagrangian approach with mass
conservation (e.g., Zubov et al., 1999), but have other
disadvantages. Hopes for improved climate simulations of
the polar regions are also associated with spherical geodesic grids, which allow for approximately uniform discretization of the sphere. Such grids are already used in
some global numerical weather-prediction models
(Majewski et al., 2002).
A key issue is the simulation of the basic physical
processes that take place in the atmosphere and determine many of the feedbacks for climate variability and
change. Examples include the representation of clouds
and radiation; dry and moist convective processes; the
formation of precipitation and its deposition on the surface as rain or snow; the interactions between the atmosphere and the land-surface orography (including the
drag on the atmosphere caused by breaking gravity
waves); and atmospheric boundary-layer processes and
their interaction with the surface. Because these processes take place on scales much smaller than the model
grid, they must be represented in terms of the largescale variables in the model (vector wind, temperature,
humidity, and surface pressure). Key atmospheric
processes from an arctic surface climate perspective
include the representations of the planetary boundary
layer, clouds, and radiation.
Energy, momentum, and moisture from the free troposphere are transferred via the atmospheric boundary layer
(ABL) to the surface and vice versa. Atmospheric general
circulation models have difficulty with the proper representation of turbulent mixing processes in general, which
Pre-release Version 24 March, 2005
www.acia.uaf.edu
106
Arctic Climate Impact Assessment
has implications for the representation of boundary-layer
clouds (IPCC, 2001).The ABL in the Arctic differs significantly from its mid-latitude counterpart, so parameterizations based on mid-latitude observations tend to perform poorly in the Arctic. Parameterizations of the surface fluxes are usually based on the Monin-Obukhov similarity theory.These parameterizations work reasonably
well for cases where the vertical stratification of the
atmosphere is weakly stable, but simulate surface fluxes
of momentum, heat, and water vapor that are too small
in the very stable stratified conditions (Poulus and Burns,
2003) common in the high Arctic. In the very stable
cases, turbulence may not be stationary, local, and continuous (Mahrt, 1998) – assumptions used in ABL parameterizations of surface fluxes. In addition, vertical resolution is a critical issue because the very thin stable surface
layer is usually shallower than the first vertical model
layer. Deviations from observations in the ABL during
winter, found in simulations with a regional climate
model for the Arctic (section 4.5.1), indicate the necessity of improvements in the atmospheric parameterization
that better describe the vertical stratification and
atmosphere–surface energy exchange (Dethloff et al.,
2001).The mean monthly turbulent heat-flux distribution at the surface strongly depends on different ABL
parameterizations and leads to different spatial distributions of temperature, wind, moisture, and other variables
throughout the arctic atmosphere.The greatest changes
are found in the ABL above the sea-ice edge in January.
Model resolution, both horizontal and vertical, is a problem in simulating the arctic ABL.The vertical discretization of current AGCMs cannot resolve the large temperature gradients and inversions that exist in the arctic ABL.
Insufficient resolution gives rise to sensible heat fluxes in
the models that tend to be too large. However, simply
increasing the resolution will not solve the problem. Even
if the very stable ABL can be simulated in finer detail, the
fundamental problem of current theories predicting turbulent fluxes that are too small will still remain.
Specific cloud types observed in the arctic ABL present a
serious challenge for atmospheric models.
Parameterizing low-level arctic clouds is particularly difficult because of complex radiative and turbulent interactions with the surface (e.g., Randall et al., 1998).
The atmospheric components of AOGCMs usually focus
on representation of tropospheric processes and the effects
of stratospheric processes on the troposphere, while their
descriptions of stratospheric processes are less satisfactory.
For example, the insufficient vertical resolution in the
stratosphere (as compared to that in the troposphere) prevents the atmospheric components of AOGCMs from
properly representing important stratospheric phenomena,
such as the quasi-biennial oscillation and sudden stratospheric temperature increases (Takahashi, 1999).
Current AOGCMs generally do not include interactive
atmospheric chemistry models (Austin et al., 2003).
Most of the atmospheric photochemical processes are
therefore simulated with chemical transport models
(CTMs) that use atmospheric wind velocities and temperature prescribed either from observational data or
from GCM simulations. In the latter case, CTMs can be
used to project the evolution of the atmospheric content
of ozone, other radiatively active gases (e.g., methane and
nitrous oxide), and aerosols (Austin et al., 2003;WMO,
2003). Projections of the distributions of tropospheric
ozone and aerosols (sulfates, soot, sea salt, and mineral
dust collectively known as “arctic haze”) are particularly
important to climate change projections (IPCC, 2001).
4.2.4. Ocean components of AOGCMs
The oceanic component in AOGCMs has improved substantially over the past decade.These models now
include representation of the full dynamics and thermodynamics of the global ocean basins and allow simulation
of the full three-dimensional current, temperature, and
salinity structure of the ocean and its evolution.
Important physical processes are associated with the
upper-ocean mixed layer and diffusive processes in the
ocean.The freezing, melting, and dynamics of sea ice
and ice–ocean interactions are also taken into account.
Until recently, because of limitations in available computing power, AOGCMs typically had similar horizontal
resolution in the ocean and atmospheric components.
Such ocean models poorly represent the large-scale
ocean current structure, not only because of the lack of
resolution of narrow boundary currents such as the Gulf
Stream and the Kuroshio, but also because of the high
viscosity coefficients necessary for computational stability (e.g., Bryan et al., 1975). However, as available computing power has increased, the resolution of the ocean
component of AOGCMs has increased to roughly one
degree of latitude and longitude. Although this resolution does not allow explicit representation of ocean
eddies (a resolution of one-third of a degree is considered “eddy permitting”, and one-ninth of a degree or
better, “eddy-resolving”), it does result in a muchimproved representation of ocean current structure.
The Arctic Ocean has always been and still remains one
of the weak spots in AOGCMs.This is partly due to specific numerical problems such as the singularity of the
longitude-latitude spherical coordinates (converging
meridians) at the North Pole (see Randall et al., 1998).
Until recently, filtering, or even inserting an artificial
island at the North Pole, have been among the usual, but
undesirable, ways to overcome the pole problem.
Rotating grids or introducing alternative grids, for example, geodesic grids providing approximately uniform discretization of the sphere (e.g., Sadourny et al., 1968) or
using curvilinear generalized coordinates (Murray, 1996),
are now being pursued in order to eliminate the converging meridian problem. Such features are now starting to
appear in oceanic components of AOGCMs (e.g., Furevik
et al., 2003). However, a greater challenge is insufficient
understanding of some phenomena related to the general
circulation of the Arctic Ocean and subarctic seas. In particular, improvement is needed in representing
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
ocean/atmosphere/sea-ice interaction processes in order
to better evaluate their importance within the context of
natural variability and anthropogenically forced change in
the climate system. A particular problem for the oceanic
component of AOGCMs is the treatment of air–ice–
ocean interactions and water-mass formation (creation of
water bodies with a homogenous distribution of temperature and salinity) over the shallow continental shelves,
which requires adequate resolution of shallow water layers, water-mass formation and mixing processes, continental runoff, and ice processes.
4.2.5. Land-surface components of AOGCMs
The land-surface components of climate models include
representation of the thermal and soil-moisture storage
properties of the land surface through modeling of its
upper layers. Key properties include surface roughness
and albedo, which are normally specified from global
datasets, although models with interactive land-surface
properties are now being developed.
Possible changes in vegetation and the effects that these
changes may have on future climate are not often taken
into account in climate change projections.These effects
may be substantial and would be manifested in the local
fluxes of water, heat, and momentum controlled by surface roughness, albedo, and surface moisture.The arctic
land types have special features that are not well represented in the present generation of climate models
(Harding et al., 2001).This is particularly true for winter
conditions where snow distribution and its interaction
with vegetation are poorly understood and modeled.
The discharge of river water to the ocean, especially to
the Arctic Ocean whose freshwater budget is much more
influenced by terrestrial water influx than are the budgets
of other oceans, is of potential importance to climate
change.The land-surface components of AOGCMs usually include river-routing schemes, in which the land surface is represented as a set of watersheds draining the
runoff (integrated over their territories at each time step)
into the grid boxes of the ocean model closest to the grid
points specified as river mouths in the land-surface
model. Such schemes are able to provide reasonable
annual means of the discharge, but shift and sharpen its
seasonal cycle, especially for the Arctic Ocean terrestrial
watersheds with their high seasonality of discharge. More
comprehensive river-routing schemes (e.g., Hagemann
and Dümenil, 1998), allowing for simulations of horizontal transport of the runoff within model watersheds, are
usually not used interactively in AOGCMs.
4.2.6. Cryospheric components of AOGCMs
Snow cover and sea ice are the two primary elements of
the cryosphere represented interactively in AOGCMs,
although some models now incorporate explicit parameterizations of permafrost processes.The large ice sheets
are represented, although non-interactively, by landsurface topography and surface albedo (typically fixed at
107
a value of around 0.8). Likewise, there is usually no
explicit representation of glaciers.
The insulating effects and change in surface albedo due
to snow cover are of particular importance for climate
change projections. AOGCMs demonstrate varying
degrees of sophistication in their snow parameterization
schemes. For example, some can represent snow density,
liquid water storage, and wind-blown snow (see Stocker
et al., 2001). Advanced albedo schemes incorporate
dependencies on snow age or temperature. However, a
major uncertainty exists regarding the ability of
AOGCMs to simulate terrestrial snow cover (McAvaney
et al., 2001; see also section 6.4), particularly its albedo
effects and the masking effects of vegetation that are
potentially important in determining the surface energy
budget (see section 7.5).
Sea-ice components of AOGCMs usually include parameterizations of the accumulation and melting of snow
on the ice, and thermodynamic energy transfers between
the ocean and atmosphere through the ice and snow.
Most of the AOGCMs evaluated by the IPCC (2001)
employed simplistic parameterizations of sea ice. Recent
advances in stand-alone sea-ice modeling, including
those in modeling sea-ice thermodynamics (e.g., introducing the effects of subgrid-scale parameterizations
with multiple thickness categories – the so-called “icethickness distribution”), are now being incorporated into
AOGCMs. However, understanding is still insufficient
for treating some atmosphere–ice–ocean interaction
issues (e.g., heat distribution between concurrent lateral
and vertical ice melt or accumulation).The primary differences among the various representations relate to
treatment of internal stresses in calculating sea-ice
model dynamics. An evaluation of the different treatment of sea-ice rheologies (relationships between internal stresses and deformation) was the core task for the
Sea-Ice Model Intercomparison Project (SIMIP) initiated
in the late 1990s. Having considered a hierarchy of
stand-alone sea-ice models with different dynamic parameterizations, SIMIP found the viscous-plastic rheology
to provide the best simulation results and adopted it as a
starting point for further optimizations (Lemke et al.,
1997). Other developments, including the elasticviscous-plastic rheology (Hunke and Dukowicz, 1997),
are helpful in achieving high computational efficiency.
However mature the status of stand-alone sea-ice
dynamics modeling, some AOGCMs still employ a simple, so-called “free drift” scheme that only allows ice to
be advected with ocean currents.There is a large range
in the ability of AOGCMs to simulate the position of the
ice edge and its seasonal cycle (McAvaney et al., 2001).
However, there is no obvious connection between the
fidelity of simulated ice extent and the inclusion of an
ice-dynamics scheme.This is apparently due to the additional impact of simulated wind-field errors (e.g., Bitz et
al., 2002;Walsh et al., 2002), which may offset
improvements from the inclusion of more realistic ice
dynamics. Conversely, the importance of improved seaice dynamics and thermodynamics has become apparent,
Pre-release Version 24 March, 2005
www.acia.uaf.edu
108
Arctic Climate Impact Assessment
and the AOGCM community is responding by including
more sophisticated treatments of sea-ice physics.
4.2.7. AOGCMs selected for the ACIA
Selecting AOGCM simulation results to be used in an
impact assessment is not a trivial task, given the variety
of models.The IPCC (McAvaney et al., 2001) concluded
that the varying sets of strengths and weaknesses that
AOGCMs display means that, at this time, no single
model can be considered “best” and it is important to
utilize results from a range of coupled models in assessment studies.The choice of AOGCMs for this assessment used the criteria suggested by Smith et al. (1998):
vintage, resolution, validity, representativeness of
results, and accessibility of the model outputs.
While models do not necessarily improve with time,
later versions (often with higher resolution) are usually
preferred to earlier ones. An important criterion for
selecting an AOGCM to be used in constructing regional
climate scenarios is its validity as evaluated by analyses of
its performance in simulating present-day and past climates (the evolution of 20th century climate in particular).The validity is evaluated by comparing the model
output with observed climate, and with output from
other models for the region of interest and larger scales,
to determine the ability of the model to simulate largescale circulation patterns.Well-established systematic
comparisons of this type are provided by international
model intercomparison projects (MIPs, see Box 4.2).
Finally, when several AOGCMs are to be selected for use
in an impact assessment, the model results should span a
representative range of changes in key variables in the
region under consideration.
Section 1.4.2 provides details of the procedure for
selecting AOGCMs for the ACIA. Initially, a set of the
most recent and comprehensive AOGCMs whose outputs were available from the IPCC Data Distribution
Center was chosen.This set was later reduced to five
AOGCMs (two European and three North American),
primarily due to the accessibility of model output, as
well as storage and network limitations. By the initial
phase of the ACIA, at least one Special Report on
Emissions Scenarios (SRES: Naki5enovi5 and Swart,
2000) B2 simulation (see section 4.4.1) extending to
2100 had been generated by each of the ACIAdesignated models. All of the models are well documented, participate in major international MIPs, and have had
their pre-SRES simulations (see Box 4.2) analyzed for
the Arctic and the results published (e.g.,Walsh et al.,
2002).The five ACIA-designated models listed in Table
4.1, together with information on their formulations,
provided the core data for constructing the ACIA climate
change scenarios.
Box 4.2. Model intercomparison projects
Model intercomparison projects (MIPs) allow comparison of the ability of different models to simulate current
and perturbed climates, in order to identify common deficiencies in the models and thus to stimulate further
investigation into possible causes of the deficiencies (Boer, 2000a,b).This is currently the only way to increase the
credibility of future climate projections. Participation in MIPs is an important prerequisite for an AOGCM to be
employed in constructing climate scenarios (e.g., for the ACIA).
In MIPs, models of the same class (AOGCMs, stand-alone AGCMs or oceanic GCMs, RCMs) are run for the same
period using the same forcings.Typically, diagnostic subprojects are established that concentrate upon analyses of
specific variables, phenomena, or regions. Occasionally, experimental subprojects are initiated, aimed mainly at
answering questions related to model sensitivity.
Of the many international MIPs conducted in the past decade, two are of primary importance for the ACIA: the
Atmospheric Model Intercomparison Project (AMIP: Gates, 1992; Gates et al., 1998), and the Coupled Model
Intercomparison Project (CMIP: Meehl et al., 2000). Both included subprojects devoted specifically to model performance at high latitudes among their numerous diagnostic subprojects.
Thirty AGCMs were included in the second phase of the AMIP (AMIP-II, concluded in 2002). All of these were
forced with the same sea-surface temperatures (SSTs) and sea-ice extents prescribed from observations, and a
set of constants, including GHG concentrations.The AMIP-II simulations span the period from 1979 to 1996. AMIP
findings related to AGCM performance in the Arctic have been reported since the early 1990s (e.g., Bitz et al.,
2002; Frei et al., 2003a; Kattsov et al., 1998, 2000;Tao et al., 1996; Walsh et al., 1998, 2002). Coupled Model
Intercomparison Project experiments belong to the class of idealized (e.g., 1% per year increase in CO2) transient
experiments with AOGCMs. Räisänen (2001) discusses some results of the second phase of the CMIP (CMIP2)
related to the Arctic (see also section 4.4.5).
The Climate of the 20th Century project was initiated in order to determine to what extent stand-alone AGCMs
are able to simulate observed climate variations of the 20th century against a background of natural variability
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
4.2.8. Summary
Atmosphere-ocean general circulation models are widely
acknowledged to be the primary tool for projecting
future climate. As understanding of the Earth’s climate
system increases and computers become more sophisticated, the scope of processes and feedbacks simulated by
AOGCMs is steadily increasing. In addition to representing the general circulation of the atmosphere and the
ocean, the AOGCMs include interactive components
representing the land surface and cryosphere.The biosphere and the carbon and sulfur cycle components of
AOGCMs are evolving, while the atmospheric chemistry
component is currently being developed off-line.The
ability to increase confidence in model projections of
arctic climate is limited by the need for further advances
in the representation of the arctic climate system in the
AOGCMs (see section 4.7).
4.3. Simulation of observed arctic climate
with the ACIA-designated models
Model-based scenarios of future climate are only credible
if the models simulate the observed climate (present-day
and past) realistically – both globally and in the region of
interest.While an accurate simulation of the present-day
climate does not guarantee a realistic sensitivity to an
external forcing (e.g., higher GHG concentrations), a
109
grossly biased present-day simulation may lead to weakening or elimination of key feedbacks in a simulation of
change, or conversely may cause key feedbacks to be
exaggerated.The ability of the models to reproduce climate states in the past – under external forcings differing
from those at present – can therefore help to add to the
credibility of their future climate projections.
Boer (2000a) distinguishes three major categories of
model evaluation: the morphology of climate, including
spatial distributions and structures of means, variances,
and other statistics of climate variables; budgets, balances, and cycles in the climate system; and process
studies of climate. A comprehensive assessment of recent
AOGCM simulations of observed global climate is provided by McAvaney et al. (2001), who, in particular,
regarded as well-established the ability of the AOGCMs
“to provide credible simulations of both the annual mean
climate and the climatological seasonal cycle over broad
continental scales for most variables of interest for climate change”. In this context, clouds and humidity were
mentioned as major sources of uncertainty, in spite of
incremental improvements in their modeling.
In this section, the first two categories of model evaluation (Boer, 2000a) are addressed for the five ACIAdesignated AOGCM simulations of the observed arctic
climate.The primary focus is on the evaluation of representations of surface air temperature and precipitation as
(Folland et al., 2002). In this MIP, the AGCMs are forced with observed SSTs and sea-ice extents and prescribed
changes in radiative forcing (GHGs, trace gases, stratospheric and tropospheric ozone, direct and indirect effects
of sulfate aerosols, solar variations, and volcanic aerosols).
The outputs of models archived at the IPCC Data Distribution Center provide an additional opportunity for
AOGCM intercomparison (IPCC-TGCIA, 1999).The archived outputs have a limited set of variables, but include
at least two scenarios (A2 and B2) from the IPCC Special Report on Emissions Scenarios (SRES: Naki$enovi$ and
Swart, 2000) and at least two pre-SRES (IS92a) emissions scenarios (GHGs only and GHGs plus sulfate aerosols).
The simulation results that are available usually span the 20th and 21st centuries.The selection of these AOGCMs
by the IPCC for use in its Third Assessment Report (IPCC, 2001) was an indication that these models provide the
most viable basis for climate change assessment.
The foci of the Arctic Regional Climate Model Intercomparison Project (ARCMIP) include the surface energy balance over ocean and land, clouds and precipitation processes, stable planetary boundary layer turbulence, icealbedo feedback, and sea-ice processes (Curry J. and Lynch, 2002; see also section 4.5.1). Another international
effort – the Arctic Ocean Model Intercomparison Project (AOMIP) – aims to identify strengths and weaknesses in
Arctic Ocean models using realistic forcing (Proshutinsky et al., 2001; see also section 4.5.2).The major goals of the
project are to examine the ability of Arctic Ocean models to simulate variability at seasonal to decadal scales, and
to qualitatively and quantitatively understand the behavior of the Arctic Ocean under changing climate forcing.
Other GCM MIPs of relevance to the ACIA include the Ocean Model Intercomparison Project (WCRP, 2002),
which is designed to stimulate the development of ocean models for climate research, and the Paleoclimate
Modeling Intercomparison Project (Braconnot, 2000), which compares AGCM/OUML models (see Box 4.1) and
AOGCMs in simulations of paleoclimate conditions during periods that were significantly different from the present-day climate.There are also a number of MIPs devoted to intercomparison of specific parameterizations
employed in GCMs, including the Sea-Ice Model Intercomparison Project (Lemke et al., 1997), the Snow Models
Intercomparison Project (Etchevers et al., 2002), and polar clouds (IGPO, 2000).
Pre-release Version 24 March, 2005
www.acia.uaf.edu
110
Arctic Climate Impact Assessment
Table 4.1. Key features of the ACIA-designated AOGCMs
Atmospheric
resolutiona
Ocean
resolutionb
Land-surface Sea-ice Flux
Primary
schemec
modeld adjustmente reference
CGCM2
Canadian Centre for Climate
Modelling and Analysis, Canada
T32 (3.8º x 3.8°)
L10
1.8° x 1.8°
L29
M, BB, F, R
T, R
H,W
Flato and Boer
(2001)
CSM_1.4
National Center for Atmospheric
Research, United States
T42 (2.8° x 2.8°)
L18
2.0° x 2.4°
L45
C, F
T, R
-
Boville et al.
(2001)
ECHAM4/OPYC3
Max-Planck Institute for
Meteorology, Germany
T42 (2.8° x 2.8°)
L19
2.8° x 2.8°
L11
M, BB, R
T, R
H*,W*
Roeckner et al.
(1996)
GFDL-R30_c
Geophysical Fluid Dynamics
Laboratory, United States
R30 (2.25° x 3.75°)
L14
2.25° x 1.875°
L18
B, R
T, F
H,W
Delworth et al.
(2002)
HadCM3
Hadley Centre for Climate Prediction
and Research, United Kingdom
2.5° x 3.75°
L19
1.25° x 1.25°
L20
C, F, R
T, F
-
Gordon et al.
(2000)
aHorizontal
resolution is expressed either as degrees latitude by longitude or as a spectral truncation (either triangular (T) or rhomboidal (R)) with a
rough translation to degrees latitude and longitude. Vertical resolution (L) is the number of vertical levels; bHorizontal resolution is expressed as
degrees latitude by longitude, while vertical resolution (L) is the number of vertical levels; cB = standard bucket hydrology scheme (single-layer reservoir of soil moisture which changes with the combined action of precipitation (snowmelt) and evaporation, and produces runoff when the water content reaches the prescribed maximum value); BB = modified bucket scheme with spatially varying soil moisture capacity and/or surface resistance; M =
multilayer temperature scheme; C = complex land-surface scheme usually including multiple soil layers for temperature and moisture, and an explicit
representation of canopy processes; F = soil freezing processes included; R = river routing of the discharge to the ocean (land surface is represented
as a set of river drainage basins); dT = thermodynamic ice model; F = “free drift” dynamics; R = ice rheology included; eH = heat flux; W = freshwater
flux; asterisks indicate annual mean flux adjustment only.
reproduced by the AOGCMs for the ACIA climatological
baseline period (1981–2000).The evaluation of individual ACIA-designated model simulations compared to historical data is also considered.
In most cases, the area between 60º and 90º N is used as
a reference region for model evaluation. In some cases,
however, smaller areas are used for consistency with
observational data (e.g., precipitation, see section
4.3.1). In cases where a variable was missing from one
of the five model outputs, a subset of four models was
evaluated for that variable.
4.3.1. Observational data and reanalyses
for model evaluation
A considerable number of datasets are available for the
Arctic, including remotely sensed and in situ data, observations from the arctic buoy program, historical data,
and field experiments (see section 2.6). However, for
evaluation of three-dimensional AOGCMs, observational
data readily available at regularly spaced grid points are
the most useful. In situ observations are not representative of conditions covering an area the size of an average
model grid box, thus a comprehensive analysis is
required to match model simulations and observations.
A good opportunity for model evaluation is provided by
reanalyses employing numerical weather prediction
models to convert irregularly spaced observational data
into complete global, gridded, and temporally homogeneous data (presently available for periods of several
decades). Reanalyses include both observed (assimilated)
variables (e.g., temperature, geopotential height) and
derived fields (e.g., precipitation, cloudiness). For some
of the derived fields, direct observations are nonexistent (e.g., evaporation).The quality of a reanalysis is
not the same for different variables; it may also vary
regionally for the same variable, depending on the availability of observations. In areas where observations are
sparse, each reanalysis primarily represents the quality of
the model’s simulation. For variables that are not
observed, the reanalysis may not be realistic. Errors in a
model’s physical parameterizations can also adversely
affect the reanalysis. However, despite these problems,
reanalyses provide the best gridded, self-consistent
datasets available for model evaluation.
It is worthwhile noting that direct point-to-point and
time-step-to-time-step comparison of a climate GCM
output against observations, reanalyses, or another climate model simulation is not methodologically correct.
Only spatial and temporal statistics can be used for the
evaluation. For state-of-the-art AOGCMs, spatial averages should be at subcontinental or greater scales, such
as the Arctic Ocean; the four ACIA regions (see section
1.1) including their marine parts; or the watersheds of
major rivers.
Observational data for validating AOGCM performance
in the Arctic (particularly the central Arctic) are characterized by a comparatively high level of uncertainty.
Because of the sparsity of direct observations, even the
temperature climatology in the Arctic is imperfectly
known. Model-simulated surface air temperature and
atmospheric pressure have primarily been compared with
Pre-release Version 24 March, 2005
www.acia.uaf.edu
111
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
NCEP/NCAR - CRU
(a)
Winter (Dec–Feb)
Spring (Mar–May)
Summer (Jun–Aug)
Autumn (Sep–Nov)
Annual Mean
Autumn (Sep–Nov)
Annual Mean
NCEP/NCAR - ECMWF
(b)
Winter (Dec–Feb)
the National Centers for
Environmental Prediction/
Spring (Mar–May)
-10
-7
Summer (Jun–Aug)
-5
-3
-1
1
3
Temperature difference (ºC)
5
7
10
large regional differences, the
NCEP/NCAR and CRU mean
Fig. 4.2. Seasonal and annual mean differences in surface air temperature (a) between the NCEP/NCAR reanalysis and the CRU
dataset for the period 1961 to 1990 and (b) between the NCEP/NCAR and the ECMWF reanalyses for the period 1979 to 1993.
National Center for Atmospheric Research (NCEP/
NCAR) reanalysis (Kistler et al., 2001).To estimate the
accuracy of the NCEP/NCAR reanalysis, its pattern of
surface air temperatures was compared against two other
datasets (Fig. 4.2).The first, compiled at the Climatic
Research Unit (CRU), University of East Anglia (New et
al., 1999, 2000), is based on the interpolation of weather
station observations. It is therefore expected to be accurate where the station density is sufficient, but it covers
only land areas.The second dataset used for comparison
is the European Centre for Medium-Range Weather
Forecasts (ECMWF) reanalysis (ERA-15; Gibson et al.,
1997). Neither of the two reanalyses should be considered as “truth” but their differences provide some information about the probable magnitude of errors in them.
The ECMWF reanalysis is only available for the period
since 1979; the difference between the ECMWF and
NCEP/NCAR reanalyses shown in Fig. 4.2 was calculated
for the overlapping interval (1979–1993).
The differences between the NCEP/NCAR reanalysis
and the CRU dataset for the period 1961 to 1990 are
smallest in summer (generally within ±1 ºC, and almost
everywhere within ±3 ºC) and largest in winter. In winter, temperatures in the NCEP/NCAR reanalysis are
higher than in the CRU dataset over most of northern
Siberia and North America, but lower over the northeastern Canadian Archipelago and Greenland. Locally,
the differences are as great as 15 ºC in northern Siberia
(NCEP/NCAR warmer than CRU) and Greenland
(NCEP/NCAR colder than CRU). Despite these very
temperatures over the entire arctic land area are in all
seasons within 2 ºC of each other.
The differences between the NCEP/NCAR and ECMWF
reanalyses over land follow the NCEP/NCAR minus
CRU differences in most, but not all, respects.
Substantial differences also occur between the ECMWF
reanalysis and the CRU dataset, most notably in spring
when the ECMWF temperatures show a widespread
cold bias compared to the CRU dataset. Over the central
Arctic Ocean, temperatures in the NCEP/NCAR
reanalysis are lower than temperatures in the ECMWF
reanalysis throughout most of the year, with the greatest
differences (up to 5–7 ºC) in autumn. In summer, however, NCEP/NCAR temperatures are slightly higher than
ECMWF temperatures.
The quality of precipitation climatologies for high latitudes derived from reanalyses is lower than that for temperature (or, e.g., atmospheric pressure). On the other
hand, assessments of simulated precipitation at high latitudes are confounded by the uncertainties in observational estimates, which suffer from errors in gauge measurements of solid precipitation, especially when the solid
precipitation occurs under windy conditions. Depending
on the partitioning between falling snow (precipitation)
and wind-blown snow from the surface, the error can
range from a significant “undercatch” to a significant
“overcatch”. Because different observational climatologies
incorporate varying types and degrees of adjustment,
there is some variance among the observational esti-
Pre-release Version 24 March, 2005
www.acia.uaf.edu
112
Arctic Climate Impact Assessment
mates, more so in the monthly means than in the annual
means (for details see sections 2.6.2.2 and 6.2.1).The
primary observational dataset used here is an outgrowth
of an arctic climatology originally compiled by Bryazgin
(1976), whose monthly mean fields were extended for
inclusion in Khrol (1996), and subsequently updated and
enhanced by additional corrections.This compilation
includes data from the Russian drifting ice stations and
high-latitude land-surface stations, and it is gridded over
the 65º to 90º N domain. Additional precipitation climatologies used in this assessment are Legates and Willmott
(1990) and Xie and Arkin (1998). Both the Bryazgin and
Legates-Willmott climatologies are based on gaugecorrected in situ data only.The two climatologies are
multi-year averages over periods that do not coincide
with each other or with the ACIA climatological baseline
(see section 4.3.2).The Xie and Arkin climatology is a
blend of in situ and satellite data, and reanalysis where in
situ and satellite data are not available. It differs significantly from the other two not only in spatial distributions, but also in areal averages.The Xie and Arkin dataset
provides monthly means for individual years over a period that includes the ACIA baseline.
4.3.2. Specifying the ACIA climatological
baseline
A climatological baseline is a period of years representing the current climate in terms of the mean and variability over the period.To satisfy widely adopted IPCC
(1994) criteria, a baseline period should:
• be representative of the present-day or recent
average climate in the region considered;
• be of sufficient duration to encompass a range of
climatic variations;
• cover a period for which data on all major climatological variables are abundant, adequately distributed in space, and readily available;
• include data of sufficiently high quality for use in
evaluating impacts; and should
• be consistent or readily comparable with baseline
climatologies used in other impact assessments.
Until recently, the most widely used baseline period has
been the “classical” 30-year period defined by the World
Meteorological Organization (WMO). Usually the period 1961 to 1990 is used (as was the case for the first
three IPCC Assessment Reports). In some cases, an earlier period (1951–1980) was used. It is expected that
the IPCC Fourth Assessment Report will use the climatological baseline 1971–2000 (IPCC-TGCIA, 1999).
The 20-year period 1981–2000 was selected as the
ACIA climatological baseline.While shorter than the
30-year WMO standard, the ACIA baseline is linked to
the period of high-quality (satellite) observations of seaice extent and concentration (important climatological
variables for the Arctic), which have been available only
since the late 1970s (IPCC, 2001; see also section 6.3).
The precise coincidence of the baseline duration with
the ACIA future time slices (also 20 years, see section
1.4.2) is also methodologically consistent. Another technical reason for selecting the 1981–2000 baseline period, rather than 1971–2000, was the availability of the
former (but not the latter) in the outputs of all five B2
simulations stored in the ACIA archive.
A serious concern is that the ACIA baseline duration is
insufficient to reflect natural climatic variability on a
multi-decadal timescale. Indeed, the ACIA climatological
baseline includes at least ten of the warmest years globally since the mid-19th century when the instrumental
record began (IPCC-TGCIA, 1999). However, considering the large interdecadal variability of arctic climate
during the entire period of the instrumental record
(e.g., Bengtsson et al., 2003; Polyakov and Johnson,
2000; Polyakov et al., 2002a,b), any particular 30-year
(or even longer) period of the 20th century could exhibit a similar limitation.
Following the recommendations of the IPCC Task Group
on Scenarios for Climate and Impact Assessment, the
ACIA climatological baseline (1981–2000) was compared
with the standard 1961–1990 baseline, using surface air
temperature and precipitation data from the
NCEP/NCAR reanalysis (Kistler et al., 2001).Table 4.2
provides seasonal and annual multi-year means of the
atmospheric variables for the two baseline periods.
The differences in the global means between the two
baseline periods are systematic, but small. Globally, the
ACIA baseline period is warmer by 0.1 to 0.2 ºC in all
seasons.The differences in global precipitation are negligible. For the polar region (60º–90º N), the differences
between the two baselines are larger.The difference in
the surface air temperature is at a maximum in winter
Table 4.2. Multi-year means of surface air temperature and precipitation derived from the NCEP/NCAR reanalysis and averaged over the WMO standard (1961–1990) and ACIA
(1981–2000) climatological baselines (Kattsov et al., 2003).
Global
Surface air
temperature (ºC)
Winter (Dec–Feb)
WMO
12.3
ACIA
12.4
Spring (Mar–May)
WMO
13.7
ACIA
13.8
Summer (Jun–Aug)
WMO
15.3
ACIA
15.5
Autumn (Sep–Nov)
WMO
13.8
ACIA
13.9
Annual
WMO
13.8
ACIA
13.9
Pre-release Version 24 March, 2005
www.acia.uaf.edu
60º–90º N
Precipitation
(mm/d)
Surface air
temperature (ºC)
2.69
2.70
-22.1
-21.5
1.10
1.12
2.71
2.72
-11.6
-10.9
1.03
1.05
2.91
2.89
5.5
5.7
1.65
1.67
2.69
2.69
-9.1
-8.6
1.31
1.31
2.75
2.75
-9.3
-8.8
1.28
1.28
Precipitation
(mm/d)
113
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Winter (Dec–Feb)
Spring (Mar–May)
-2
-1.5
Summer (Jun–Aug)
-1
-0.5
0
0.5
Temperature difference (ºC)
Autumn (Sep–Nov)
1
1.5
Annual Mean
2
Fig. 4.3. Differences in seasonal and annual multi-year mean surface air temperature in the northern polar region (60º–90º N) between
the ACIA (1981–2000) and the standard WMO (1961–1990) climatological baselines, obtained from the NCEP/NCAR reanalysis.
(0.6 ºC) and is smallest in summer (0.2 ºC).The ACIA
baseline annual mean precipitation is the same as the
1961–1990 mean.
Geographically, the differences between the two baseline
periods in the NCEP/NCAR reanalysis are more pronounced (Fig. 4.3).The Arctic is generally warmer during the ACIA baseline period, especially in autumn.The
strongest warming is evidently associated with the marginal sea-ice zone, particularly along the east coast of
Greenland. Mean sea-level pressure (SLP) is generally
lower (by up to about 1.5 hPa) over the central Arctic
and northern North Atlantic in the ACIA baseline period
compared to the 1961–1990 baseline period.
Because 1961–1990 is expected to be superseded in the
near future by 1971–2000 as the new standard 30-year
averaging period, it is worth comparing the ACIA climatological baseline against the latter. Figure 4.4 shows the
spatial distributions of seasonal and annual differences in
surface air temperature between the 1981–2000 and
1971–2000 periods.
In summary, for surface air temperature, precipitation,
and atmospheric pressure, the ACIA baseline period has
systematic but generally small differences in comparison
with the WMO standard baseline (1961–1990).The differences can easily be taken into account when a comparison between climate change scenarios employing the
Winter (Dec–Feb)
Spring (Mar–May)
-2
-1.5
different baselines is required. An advantage of the ACIA
climatological baseline period is that it is more “current”
than the 1961–1990 period.The duration of the ACIA
baseline period is exactly the same as that adopted for
the ACIA future time slices.There are only minor geographical differences in seasonal temperature means
between the ACIA baseline and the new standard baseline (1971–2000).While the ACIA climatological baseline period (1981–2000) satisfies the IPCC (1994) selection criteria, its relative shortness (compared to the
standard 30-year period) may not provide an adequate
representation of the probability of extreme events.
4.3.3. Surface air temperature
The seasonal cycle of the simulated and analyzed air
temperatures at the 2 m height (1.5 m height for
HadCM3) for the period 1981 to 2000 is illustrated in
Fig. 4.5a, which shows means for the entire area north
of 65º N.While there are differences between the fivemodel mean and the NCEP/NCAR reanalysis, these are
relatively small compared with the range of model
results. In particular, the area mean temperatures during
the greater part of the year differ by about 5 ºC between
the models simulating the highest and lowest temperatures, and local differences are even larger. In late winter
(February–March), all the models simulate slightly lower
temperatures than those in the NCEP/NCAR reanalysis.
It should be noted that the 1.5 and 2 m temperatures in
Summer (Jun–Aug)
-1
-0.5
0
0.5
Temperature difference (ºC)
Autumn (Sep–Nov)
1
1.5
Annual Mean
2
Fig. 4.4. Differences in seasonal and annual multi-year mean surface air temperature in the northern polar region (60º–90º N) between
the ACIA (1981–2000) and the new standard (1971–2000) climatological baselines, obtained from the NCEP/NCAR reanalysis.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
114
Arctic Climate Impact Assessment
America (Fig. 4.6). A sharp local maximum in the fivemodel bias (Fig. 4.6c) in southern Greenland probably
reflects the relatively smooth model topographies.
However, the biases vary substantially between the individual models. Figure 4.6d shows the number of the
five models that simulate lower mean annual temperatures than those in the NCEP/NCAR reanalysis.There
are a few areas in the western Arctic where all five
models simulate higher mean annual temperatures than
those in the NCEP reanalysis and a few areas in the
eastern Arctic where all five models simulate lower
mean annual temperatures.The seasonal distribution of
the biases in the five-model mean temperature is shown
in Fig. 4.6e-h.The cold bias in northern Eurasia is most
pronounced in winter and spring, whereas the warm
bias in northern North America persists for most of the
year and is largest in autumn.The simulated temperatures in the central Arctic Ocean tend to exceed the
NCEP/NCAR reanalysis estimate in spring and especially in autumn.The five-model mean simulated summer
temperatures in the central Arctic are lower than in the
NCEP/NCAR reanalysis and the model-to-model variation is relatively small.
Fig. 4.5. Seasonal cycles of (a) surface air temperature and (b)
precipitation simulated by the five ACIA-designated models for
the period 1981–2000 and averaged over the area 65º–90º N.
For comparison, data for the same area are included in (a) from
the NCEP/NCAR reanalysis for the same time period and in (b)
from three climatologies: Bryazgin (1936–1990), LegatesWillmott (1920–1980), and Xie-Arkin (1981–2000) (Khrol, 1996;
Legates and Willmott, 1990; Xie and Arkin, 1998).
the models are not prognostic variables – they are
derived from the prognostic temperatures at the lowest
model level (typically a few tens of meters) using the
models’ ABL parameterizations; as a result, the biases
may be partly due to the diagnostic schemes. In addition, surface elevation differences between the models
and the reanalysis, as a result of differences in spatial resolution, could be contributing to the apparent biases.
The large-scale spatial distribution of annual mean temperature in the Arctic is, on average, reasonably well
reproduced by the ACIA-designated models.The simulations tend to be slightly colder than the NCEP/NCAR
reanalysis in northern Eurasia and somewhat warmer
over the western Arctic Ocean and northern North
The ability of AOGCMs to provide credible projections
of future climates is strongly supported by their ability
to simulate the evolution of the climate during past centuries. One of the five-member ensemble 20th century
simulations with the GFDL-R30_c model driven by historical changes in GHG and sulfate aerosol concentrations demonstrated an impressive resemblance to the
observed warming that occurred in two distinct periods
in the first and second halves of the 20th century with a
pronounced maximum in the Arctic (Delworth and
Knutson, 2000).The early 20th-century warming was
not obtained in the other simulations of the GFDLR30_c ensemble, which highlights the role of internal
variability in the climate evolution, and therefore proves
the necessity of ensemble simulations, rather than single
runs, in order to better delineate the associated uncertainties. In the HadCM3 model, a good fit to 20th century observations was only obtained when natural (varying) forcing from solar and volcanic activity was also
included (Stott et al., 2000; Stott, 2003;Tett et al.,
2000; see also IPCC, 2001a).
4.3.4. Precipitation
More so than with temperature, there are major systematic differences between precipitation in the ACIA
1981–2000 simulations and in observational datasets.
The five-model mean seasonal cycle of precipitation in
the area 65º to 90º N is in qualitative agreement with
the climatologies (Fig. 4.5b).While the range between
the individual simulations is substantial throughout the
year, it is noteworthy that the observational climatologies demonstrate a comparable scatter, at least in summer and autumn.
As shown in Fig. 4.7a-d, the average simulated annual precipitation generally exceeds the Bryazgin estimate
Pre-release Version 24 March, 2005
www.acia.uaf.edu
115
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
(a) 5 Models, Annual
-30
-25
(b) NCEP, Annual
-20 -15 -10
-5
0
Surface air temperature (ºC)
(e) Bias, Dec–Feb
5
-5
-3 -1 1
3
5
Temperature difference (ºC)
(f) Bias, Mar–May
(Bryazgin, 1976; Khrol, 1996) and other
observational estimates in most of the Arctic;
(c) Bias, Annual
(g) Bias, Jun–Aug
-5
-3 -1 1
3
5
Temperature difference (ºC)
(d) N Models < NCEP
0
1
2
3
4
Number of models
5
(h) Bias, Sep–Nov
and Walsh, 2002). In all of the ACIA-designated model simulations, positive linear
Fig. 4.6. Comparison of surface air temperature simulated by the ACIA-designated models for 1981–2000 and the NCEP/NCAR
reanalysis temperature for the same period.The top row shows (a) the simulated annual mean temperature averaged over the five
ACIA-designated models; (b) the NCEP/NCAR reanalysis for 1981–2000; (c) the difference between (a) and (b); and (d) the number of models (out of five) in which the simulated annual mean temperature is lower than that of the NCEP/NCAR reanalysis.The
bottom row shows seasonal differences between the five-model mean and the NCEP/NCAR reanalysis for (e) winter; (f) spring; (g)
summer; and (h) autumn.
trends in 20th century arctic precipitation agree with
available observational estimates (see section 2.6.2.2).
in some areas by a factor of two.The reverse is true, however, in the northeastern North Atlantic and parts of
northwestern Eurasia, probably because simulated cyclone
activity in this area tends to be too weak (see section
4.3.5).The same geographical pattern of biases persists
throughout the year, although the magnitude of these biases varies with season.The positive biases relative to the
Bryazgin climatology are generally greatest in spring and
smallest in summer (Fig. 4.7e-h).
4.3.5. Other climatic variables
The differences between the simulated precipitation and
the observational estimates may be partly due to measurement errors that lead to underestimation of the actual precipitation, particularly when it falls in solid form.
However, this clearly cannot explain all the differences
in the spatial and seasonal distributions of precipitation.
For example, the difference between the five-model area
mean and the observational estimates is substantially
larger in spring than in winter, in contrast to what might
be expected from measurement errors alone.
The distribution of the 1981–2000 mean SLP is qualitatively similar between the mean of four of the ACIA-designated models (for GFDL-R30_c, only surface pressure
fields were available) and the NCEP/NCAR reanalysis, but
there are important differences in details (Fig. 4.8). Surface
pressure is the prognostic variable in GCMs, while SLP is
diagnosed using different reduction schemes. Some of the
variations in SLP between the different models, and
between the models and the reanalysis, may be due to the
use of different SLP reduction schemes.
The ability of AOGCMs (including three of the ACIAdesignated models: HadCM3, ECHAM4/OPYC3, and
CGCM2) to reproduce the 20th century increase in arctic precipitation has been demonstrated (e.g., Kattsov
The models simulate the main lobes of the annual mean
Icelandic and Aleutian lows quite well, but the simulated
extension of the Icelandic low towards the Barents Sea is
too weak.The positive pressure bias over the eastern
Further details on the evaluation of the five ACIA-designated models with respect to simulation of precipitation
and other components of arctic hydrology and climatology can be found in Chapter 6.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
116
Arctic Climate Impact Assessment
(a) 5 Models, Annual
0.5
(b) Bryazgin, Annual
1
1.5
2
3
Precipitation (mm/d)
(e) Bias, Dec–Feb
(c) Bias, Annual
-50 -30 -10 10 30 60 100
Difference (%)
(f) Bias, Mar–May
(g) Bias, Jun–Aug
(d) N Models > Bryazgin
0
1
2
3
4
Number of models
5
(h) Bias, Sep–Nov
-50 -30 -10 10 30 60 100
Difference (%)
Fig. 4.7. Comparison of precipitation simulated by the ACIA-designated models for 1981–2000 and precipitation from the Bryazgin
climatology (Khrol, 1996) for 1936–1990.The top row shows (a) the simulated annual mean precipitation averaged over the five
ACIA-designated models; (b) the Bryazgin climatology for 1936–1990; (c) the percentage difference between (a) and (b); and (d) the
number of models (out of five) in which the simulated annual mean precipitation exceeds that of the Bryazgin climatology.The bottom row shows seasonal percentage differences between the five-model mean and the Bryazgin climatology for (e) winter; (f)
spring; (g) summer; and (h) autumn.
Arctic Ocean and the negative bias in western Eurasia
(Fig. 4.8c) suggest that the path of cyclone activity is too
far south in the simulations.There also tends to be a
slight negative pressure bias in the western Canadian
Arctic, which suggests that the simulated cyclone activity
is too strong in that region.Winter pressure biases make
the greatest contribution to the four-model annual mean
pressure biases. However, the positive bias over the eastern Arctic Ocean persists throughout the year. As with
temperature and precipitation, the pressure biases also
vary between the individual models.The shift of the arctic air mass relative to observations is a well-known feature of both AOGCMs and stand-alone AGCMs (e.g.,
AMIP; see Box 4.2).The pressure biases contribute to
significant differences in wind forcing of sea ice between
the AOGCM simulations and the real world, and lead to
distortions in simulated spatial distributions of sea ice
(see Bitz et al., 2002;Walsh, in press;Walsh et al., 2002).
Cloudiness and the radiative properties of clouds, particularly in the Arctic, remain a major challenge to simulate.
Figure 4.9 shows the dramatic scatter between the total
cloud amounts simulated by four of the ACIA-designated
models (the results from CGCM2 were not available) over
the Arctic Ocean between 70º and 90º N.Two observational estimates, one based primarily on Television and
Infrared Observation Satellite Operational Vertical
Sounder (TOVS) data (Schweiger et al., 1999) and the
other obtained primarily from surface-based observations
(Hahn et al., 1995), diverge substantially in late summer
and early autumn, but give an idea of the seasonality of
arctic cloud cover.While the inter-model scatter is quite
large (e.g., the difference between the highest (CSM_1.4)
and lowest (GFDL-R30_c) simulations approaches 60% in
winter), the four-model mean underestimates cloudiness
in the warm season and overestimates it in winter. Of the
four ACIA-designated models, only the HadCM3 simulation shows some qualitative agreement with the observed
seasonality of arctic cloud cover.
Surface radiative fluxes also vary widely between the
models. Figure 4.10 shows the seasonal variation in incident solar radiation as simulated by four of the ACIAdesignated models (ECHAM4/OPYC3 data were not
available) for the Arctic Ocean between 70º and 90º N
and the lone observational estimate obtained from the
Langley Atmospheric Sciences Data Center.The fourmodel mean seasonal cycle is close to the observed one.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
117
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
(a) 4 Models, Annual
(b) NCEP, Annual
1005 1010 1015 1020
Sea-level pressure (hPa)
(e) Bias, Dec–Feb
(c) Bias, Annual
-6 -5 -4 -3 -2 -1 1 2 3 4 5 6
Pressure difference (hPa)
(f) Bias, Mar–May
(g) Bias, Jun–Aug
(d) N Models > NCEP
0
1
2
3
Number of Models
4
(h) Bias, Sep–Nov
-6 -5 -4 -3 -2 -1 1 2 3 4 5 6
Pressure difference (hPa)
Fig. 4.8. Comparison of mean sea-level pressure (SLP) simulated by four of the ACIA-designated models for 1981–2000 and the
NCEP/NCAR reanalysis SLP for the same period.The top row shows (a) the simulated annual mean SLP averaged over the four
ACIA-designated models; (b) the NCEP reanalysis for 1981–2000; (c) the difference between (a) and (b); and (d) the number of
models (out of four) in which the simulated annual mean SLP exceeds the NCEP/NCAR reanalysis.The bottom row shows seasonal differences between the four-model mean and the NCEP/NCAR reanalysis for (e) winter; (f) spring; (g) summer; and (h) autumn.
In summer, however, the difference between the highest
and lowest simulated values (CGCM2 and CSM_1.4,
respectively) reaches a maximum of up to 125 W/m2.
The large inter-model scatter in the ACIA-designated
model simulations of baseline (1981–2000) sea-ice and
terrestrial snow-cover distributions reflects problems in
modeling the cryosphere as discussed in sections 4.2.5
and 4.2.6. Figure 4.11 presents an integrated picture of
sea-ice distributions in the Northern Hemisphere as simulated by the five ACIA-designated models for March
(maximum observed sea-ice extent) and September
(minimum extent).The distribution is represented by the
number of models simulating sea ice in each of the 2.5º x
2.5º grid cells. For each model, sea ice is defined as present in a grid cell if its quantity exceeds one of the ad hoc
critical values (depending on what variable is available
from the model output): 5 cm thickness, 45 kg/m2 mass,
or 10% areal coverage. For the greater part of the Arctic
Ocean, all five models simulate sea ice in both March and
September; however, major differences between the
models occur along the margins of the ice cover.
A detailed evaluation of the five ACIA-designated models
with respect to simulation of the baseline sea-ice and
terrestrial snow-cover distributions in the Northern
Hemisphere is provided in sections 6.3.3 and 6.4.3.
4.3.6. Summary
A key characteristic of the simulations of the arctic climate from the ACIA-designated models is their large
inter-model scatter. Biases in surface air temperature and
SLP spatial distributions and simulation of excessive arctic precipitation are among the most important systematic errors. Significant uncertainty is introduced by the
simulated cloudiness – one of the key variables in climate system feedbacks (this is also a problem in the current generation of AOGCMs outside of the Arctic).The
large inter-model scatter in reproducing sea-ice and terrestrial snow-cover extent limits the credibility of future
climate projections obtained with the models.
Conversely, compared to the five individual simulations,
the five-model ensemble means show reasonable agreement with available observations, at least for the area
averages.The evaluation of the ability of the ACIAdesignated AOGCM ensemble to simulate observed climate conditions supports use of the ensemble for constructing 21st-century climate change scenarios for the
Arctic.This suitability is further supported by the ability
Pre-release Version 24 March, 2005
www.acia.uaf.edu
118
Arctic Climate Impact Assessment
Fig. 4.9. Annual cycle of monthly mean cloudiness over the
Arctic Ocean (70º–90º N) for 1981–2000 simulated by four of
the ACIA-designated models, and observational estimates from
TOVS satellite data and surface observations (Hahn et al., 1995).
Fig. 4.10. Annual cycle of incident solar radiation at the surface
of the Arctic Ocean (70º–90º N) for 1981–2000 simulated by
four of the ACIA-designated models, and the observationally
based estimate obtained using data from the Langley
Atmospheric Sciences Data Center (1983–1991).
of some of the ACIA-designated models, when driven by
estimates of historical radiative forcing, to satisfactorily
emulate the observed evolution of arctic surface air temperature and precipitation throughout the 20th century,
which enhances the credibility of future arctic climate
change projected by these models.
of 3.0 ºC (with a range of 1.3–4.5 ºC derived from the
nine models used by the IPCC) for the A2 emissions scenario and 2.2 ºC (with a range of 0.9–3.4 ºC) for the B2
emissions scenario. Most of the spatial patterns of the
AOGCM-projected responses to the SRES emissions
scenarios are similar to other emissions scenarios,
including the idealized 1% per year CO2 increase.The
following list summarizes the key projections of climate
change over the 21st century by the AOGCMs used in
the IPCC (2001) assessment.
As a consequence of the biases and inter-model scatter in
AOGCM simulations of the present-day arctic climate,
the ACIA has chosen to append (add or subtract) the simulated changes to observed baseline climates as was done,
for example, by the National Assessment Synthesis Team
(NAST, 2001) rather than to use the model-simulated climates directly in impact studies.The climate changes
should be expressed either as absolute differences (e.g.,
temperature), or as ratios (e.g., precipitation).
4.4. Arctic climate change scenarios for
the 21st century projected by the
ACIA-designated models
The IPCC Third Assessment Report (IPCC, 2001), based
on a set of AOGCM projections, has provided the following global context for the ACIA. For the last three
decades of the 21st century (2071–2100), the IPCC
(2001) projects a mean increase in globally averaged surface air temperature, relative to the period 1961–1990,
(a)
March
(b)
• It is very likely that nearly all land areas will warm
more rapidly than the global average, particularly
during the cold season in northern high latitudes.
• Models project a decrease in the diurnal temperature range in many areas, with nighttime lows
increasing more than daytime highs.
• Models project a decrease in Northern
Hemisphere snow cover and sea-ice extent, and
continued retreat of glaciers and ice caps.
• Projected increases in mean precipitation are likely
to lead to increases in interannual precipitation
variability.
• Increases in the lowest daily minimum temperatures are projected to occur over nearly all land
areas and are generally greatest in areas where
snow and ice retreat.
September
Number of
Models
1
2
3
4
5
Observed
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Fig. 4.11. Baseline
(1981–2000) sea-ice distributions in the Northern
Hemisphere simulated by
the ACIA-designated models for (a) March (maximum observed sea-ice
extent) and (b) September
(minimum extent) in terms
of the number of models
simulating sea ice in each
grid cell (based on
Meleshko et al., 2004).
119
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Box 4.3. Uncertainties in climate change scenarios based on AOGCM simulations
Uncertainties in future GHG and aerosol emissions, their conversion to atmospheric concentrations, and their contribution to
radiative forcing of the climate. Different assumptions about future social and economic development, and hence
future GHG and aerosol emissions, comprise one of the major uncertainties in the climate change scenarios. For
example, the IPCC Special Report on Emissions Scenarios (Naki$enovi$ and Swart, 2000; see also section 4.4.1)
presents 40 different emissions scenarios. Uncertainty is also associated with the conversion of emissions into atmospheric GHG and aerosol concentrations. Additional uncertainty arises from the calculation of radiative forcing associated with given concentrations, which occurs implicitly within AOGCMs, but is problematic in particular for aerosols.
Uncertainties in the global and regional climate responses to a radiative forcing from different AOGCM simulations. Due
to different representations of processes and feedbacks in the climate system (e.g., Stocker et al., 2001), or by
excluding some of them, AOGCMs differ in their sensitivity to the same radiative forcing. Sometimes, the difference in AOGCM sensitivity manifests itself only in projected regional climate change patterns, while the magnitudes of projected global mean changes remain similar between models. At long timescales, the forcing uncertainty (differences between emissions scenarios) and model uncertainty are of approximately equal importance.
Uncertainties due to insufficient AOGCM resolution and different methods of regionalizing (downscaling) AOGCM results.
Insufficient AOGCM resolution limits direct use of their outputs in impact assessments. In most cases, a climate scenario for a certain region requires a combination of the simulated variables and observed data, which may be
accomplished with different methods. In addition, observational data often fail to capture the full range of decadalscale natural variability. Finally, gridding the observational data in order to create baseline climatologies can introduce
errors. Employing regional climate models to enhance the spatial and temporal resolution of AOGCM outputs introduces further uncertainties arising from individual features of the RCMs. In principle, by employing a range of downscaling methods, quantification of this class of uncertainties is possible, but this is seldom done (Mearns et al., 2001).
Uncertainties due to forced and unforced natural variability. In addition to anthropogenic forcing, climate change in
the real world is affected by largely unpredictable natural variability. Part of the natural variability is thought to be
due to variations in solar and volcanic activity, but a substantial part is unforced, resulting from the internal dynamics of the climate system. Climate models also simulate unforced natural variability, such that, when the same
model is run with the same forcing scenario but different initial conditions, there are non-negligible differences in
the results, particularly in regional details that are affected by internal variability much more than global means.
When different models are run using the same forcing scenario, some of the differences in their results arise from
different realizations of natural variability.
• Frost days and cold waves are very likely to
become fewer.
• High extremes of precipitation are projected to
increase more than the mean, and the intensity of
precipitation events is projected to increase.
• The frequency of extreme precipitation events is
projected to increase almost everywhere.
• For some other extreme phenomena, many of
which may have important impacts on the environment and society, the confidence in model projections and understanding is currently inadequate to
make firm projections.
• No clear consensus has been reached about how
extratropical cyclones are likely to change, as the
results differ between the relatively few studies
that have been conducted.
• Most AOGCMs project a weakening of the
Northern Hemisphere thermohaline circulation,
which would contribute to a reduction of surface
warming in the subarctic North Atlantic.
• There is no clear agreement on likely changes in
the probability distribution or structure of natural
modes of variability, like the North Atlantic
Oscillation (NAO) or the Arctic Oscillation (AO),
whose magnitude and character changes vary
across the models.
Box 4.3 reviews some of the major uncertainties associated with AOGCM projections.
In this section, the Arctic is considered in the context of
the 21st-century global climate change projections listed
previously, focusing on surface air temperature and precipitation in the Arctic and the inter-model differences
for the A2 and B2 emissions scenarios. All changes are
compared to the ACIA climatological baseline (1981–
2000). A wider context of climate change simulations is
provided by comparing the behavior of the five ACIAdesignated AOGCMs in Phase 2 of the Coupled Model
Intercomparison Project (CMIP2: Meehl et al., 2000)
with the other 14 models included in that project.
4.4.1. Emissions scenarios
Emissions scenarios are plausible representations of the
future development of emissions of radiatively active
substances (GHGs, aerosols), based on a coherent and
internally consistent set of assumptions about demo-
Pre-release Version 24 March, 2005
www.acia.uaf.edu
120
Arctic Climate Impact Assessment
Fig. 4.12. Comparison of the six SRES marker scenarios and the IS92a scenario showing (a) CO2 emissions; (b) the resulting atmospheric CO2 concentrations; (c) sulfur dioxide (SO2) emissions; and (d) projections of global mean temperature increases relative to
1990. In (d), the projected temperature increase for each emissions scenario is an average of the results of the seven climate models
used by IPCC (2001).The dark shading represents the range of average warming across all 40 SRES scenarios, and the light shading the
range across all scenarios and models (based on IPCC, 2001).
graphic, socioeconomic, and technological changes and
their key relationships in the future. Emissions scenarios
are converted into concentration scenarios that are used
as input for climate model projections.
In idealized transient experiments (e.g., CMIP2, see Box
4.2), the atmospheric CO2 concentration increases gradually, usually at a rate of 1% (compound) per year,
which results in a doubling of its concentration in 70
years.The 1% idealized scenario lies at the high end of
the SRES scenarios.
In transient experiments with a detailed forcing scenario,
the concentrations of CO2 and other GHGs such as
methane and nitrous oxide are prescribed as a function of
time, based on an emissions scenario for these gases.
Frequently, sulfate aerosols are also included. Examples
of the scenarios used in model simulations include the
IPCC IS92 scenarios (Leggett et al., 1992) and the more
recent SRES scenarios (Naki5enovi5 and Swart, 2000).
The SRES emissions scenarios were built around four
narrative storylines that describe the evolution of the
world in the 21st century. Altogether, 40 different emis-
sions scenarios were constructed. Six of these (A1B,
A1T, A1FI, A2, B1, and B2) were chosen by the IPCC as
illustrative “marker” scenarios.The SRES scenarios
include no additional mitigation initiatives, which means
that no scenarios are included that explicitly assume the
implementation of the United Nations Framework
Convention on Climate Change or the emission targets
of the Kyoto Protocol.
The greatest difference between the SRES scenarios and
the earlier IS92 scenarios relates to sulfur emissions and
hence sulfate aerosol concentrations.The commonly
adopted intermediate IS92 scenario (IS92a) assumed a
doubling of anthropogenic sulfur emissions between
1990 and 2050, and little change in emissions thereafter.
In contrast, all six illustrative SRES scenarios project
lower sulfur emissions in 2100 than at present, although
some include an increase over the next few decades.The
lower sulfur emissions together with higher GHG emissions in some of the SRES scenarios are the main reason
for the upward shift in the IPCC projections of the
increase in global mean temperature between 1990 and
2100 from 1.0 to 3.5 ºC in the Second Assessment
Report (IPCC, 1996) to 1.4 to 5.8 ºC in the Third
Pre-release Version 24 March, 2005
www.acia.uaf.edu
121
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Assessment Report (IPCC, 2001).The CO2 emissions
and the derived atmospheric CO2 concentrations, the
sulfur dioxide (SO2) emissions, and projections of global
mean temperature increases for the SRES marker scenarios and for the IS92a scenario are shown in Fig. 4.12.
Box 4.4 discusses specific issues related to the use of the
ACIA-designated model projections as a basis for local
scenarios.
No probabilities are assigned to the various SRES scenarios. During the initial stage of the ACIA process, to stay
coordinated with current IPCC efforts, it was agreed
that the ACIA projections would be based on the IPCC
SRES scenarios (Källén et al., 2001). By that time, most
of the available (and expected to be shortly available)
AOGCM simulations to be relied upon had been forced
by two emissions scenarios: A2 and B2 (Cubasch et al.,
2001). Globally, the model mean transient climate
responses to the A2 and B2 emissions scenarios are close
to each other for each of the different models through
the first half of the 21st century and only diverge significantly after that. Given the schedule for producing this
assessment and the limits of resources for data storage,
the B2 emissions scenario was chosen as the primary
scenario for use in ACIA impact analyses.
Figure 4.13 displays the evolution of arctic annual mean
temperature during the 21st century projected by the
five ACIA-designated model simulations using the B2
emissions scenario.The projections for the 60º to 90º N
polar region are shown along with the range of global
mean temperature changes projected by the same simulations. In the five ACIA-designated model projections,
by the late 21st century (2071–2090), the global mean
temperature increase (from the 1981–2000 baseline)
varies from 1.4 ºC (CSM_1.4) to 2.1 ºC (ECHAM4/
OPYC3 and CGCM2), with a five-model average of
1.9 ºC. In the Arctic, the increase in mean annual temperature projected by the five models is significantly
larger, reaching 3.7 ºC (twice the increase in the global
mean) for the area north of 60º N.The projected temperature change varies by about ±25% about the mean
of the five models. For the Arctic north of 60º N, the
projected area mean temperature increase by 2071–
2090 ranges from 2.8 ºC (CSM_1.4) to 4.6 ºC
(ECHAM4/OPYC3), with the other models within the
3.7 to 4.0 ºC range (Table 4.3).The projected arctic
mean temperature increase exceeds the projected global
mean temperature increase in all of the models.
In a number of studies (e.g., Carter et al., 2000;
Ruosteenoja et al., 2003), a “pattern-scaling” technique
is applied to represent a wider range of possible future
forcings than are available from AOGCM simulations
alone. In particular, Ruosteenoja et al. (2003) extrapolated available A2 and B2 AOGCM projections of temperature and precipitation for the SRES B1 and A1FI scenarios over 32 world regions, including the Arctic.The
application of this technique is based on the assumption
that the spatial pattern of the response is independent of
the forcing, while the amplitude of the response at each
location is linearly proportional to the global mean
change in surface air temperature.The global mean temperature changes for the entire range of SRES scenarios
(as shown by shading in Fig. 4.12d) were calculated
using a simple climate model system (Box 4.1) calibrated to be consistent with each AOGCM being emulated.
Additional assumptions of this approach, first suggested
by Santer et al. (1990), are that the patterns of the climate response to anthropogenic forcing can be adequately defined from AOGCM simulations and that they
are stable through time and across a representative range
of possible anthropogenic forcings. Uncertainties due to
scaling climate response patterns increase for scenarios
that include substantial, regionally differentiated aerosol
forcings and in regions where there is an enhanced
response, for example, near sea-ice and snow margins
(Carter et al., 2000).
4.4.2. Changes in surface air temperature
In Fig. 4.14, spatial patterns of projected increases in the
annual mean temperature in the Arctic are put in a global perspective for the three 21st-century time slices. On
average, the models project a greater temperature
increase at high northern latitudes than anywhere else in
the world (Fig. 4.14, left column). By 2071–2090, the
five-model average projected increase in the mean annual temperature in the central Arctic is more than 5 ºC
(about three times the global mean). By that time, the
mean annual temperature is projected to increase by
around 3 ºC in Scandinavia and East Greenland, about
2 ºC in Iceland, and up to 5 ºC in the Canadian
Archipelago and Russian Arctic. However, the variation
in projected temperature change between the individual
models is also generally much larger over the Arctic than
for most other regions of the globe (Fig. 4.14, middle
column).The standard deviation of the mean temperature change averaged over the five models also varies
substantially across the Arctic, but these variations are
difficult to interpret because there are only five models
in the sample.The relative agreement between the different projections is measured by the ratio between the
Table 4.3. Increases in mean annual surface air temperature in the Arctic (60º–90º N) compared to the 1981–2000 baseline, as projected by the five ACIA-designated models forced with the B2 emissions scenario (Kattsov et al., 2003).
Temperature change (ºC)
2011–2030
2041–2060
2071–2090
CGCM2
1.2
2.5
3.7
CSM_1.4
1.5
2.2
2.8
ECHAM4/OPYC3
1.3
3.2
4.6
GFDL-R30_c
1.0
2.5
3.8
Pre-release Version 24 March, 2005
www.acia.uaf.edu
HadCM3
1.1
2.2
4.0
Five-model mean
1.2
2.5
3.7
122
Arctic Climate Impact Assessment
Box 4.4. Specific issues related to the use of ACIA-designated AOGCM projections as a basis
for local scenarios
There are two basic approaches for applying climate change projections in impact studies: deterministic and probabilistic. In the deterministic approach, a single best-guess projection of climate change is used for impact modeling.The simplest way of making a best-guess projection, which is based on the assumption that all models give
equally likely results, is to use the arithmetic mean of all available model results (such as the mean of the five
ACIA simulations). In the probabilistic approach, in contrast, the projections from different climate models are
treated as giving a probability distribution of future climate changes (Giorgi and Mearns, 2003; Palmer and
Räisänen, 2002; Räisänen and Palmer, 2001). With this approach, each climate model projection is used separately
in an impact analysis. If averaging of different scenarios is performed, it is done in the last phase: the calculated
impacts, rather than the climate change projections, are averaged. Because the impacts may depend nonlinearly on
changes in climate, the average impact scenario derived by the probabilistic method may differ from the impact
scenario obtained by using the average climate change projection. For this reason and because information on
uncertainty of the impacts is also important, the probabilistic approach is, in principle, preferable to the deterministic approach. However, it is also more demanding in terms of the computations required.
The simplest variants of both deterministic and probabilistic methods assume that all available climate model
results are equally likely. In some situations, this simple assumption may be questionable. For example, when models have serious problems in their control climates, it may be best to exclude them from the calculations.
Furthermore, for some situations different models may give such widely divergent results that a deterministic
averaging may be misleading. Examples of situations in which local scenario construction requires special care
include the following.
• Climate changes in the North Atlantic depend to a high degree on the state of the ocean circulation. Some
models project a significantly reduced thermohaline circulation as a consequence of climate change, while oth-
five-model mean change and the inter-projection standard deviation (Fig. 4.14, right column). Because the
large standard deviations compensate for the large average warming, this signal-to-noise ratio in the Arctic is
not exceptional.Very high relative agreement (mean
exceeding the standard deviation by a factor of six)
occurs by 2071–2090 in some (but not all) arctic
regions, but this is also the case in lower latitudes where
both the mean and the standard deviation are smaller.
The differences in temperature change between the five
ACIA-designated simulations are caused primarily by
model differences and by noise associated with internal
variability.To investigate the role of the latter factor
alone, the five-model mean temperature changes were
compared with the variability of 20-year mean temperatures generated in CMIP2 control simulations.The signalto-noise ratio that was estimated by dividing the average
warming by √2 times the standard deviation of 20-year
means is shown in Fig. 4.15.The ratios are generally
higher than those in Fig. 4.14, because internal variability
does not cause all differences between the ACIAdesignated simulations. Importantly, however, the ratio is
still relatively low (<2) in some parts of the Arctic in the
period 2011–2030. Because a ratio of two defines the
lower limit of statistical significance, this suggests that, at
least in some parts of the Arctic, the GHG-induced temperature increase may remain difficult to differentiate
from natural variability over the next few decades.
Indeed, the real signal-to-noise ratio might be even lower
than calculated here, because the CMIP2 simulations
exclude external sources of natural variability. Later in
the 21st century, the signal-to-noise ratio increases, but
remains generally lower in the Arctic than at low latitudes, where the internal variability is much smaller.
Thus, despite the large average warming suggested by
these simulations, the Arctic might not be the area where
anthropogenic climate changes are easiest to detect.
Fig. 4.13. Global and arctic (60º–90º N) changes in annual mean
surface air temperature relative to the baseline period
1981–2000 as projected by the five ACIA-designated models
forced with the B2 emissions scenario.
Together, the results in Figs. 4.14 and 4.15 emphasize
the importance of improving understanding of how best
to interpret the relatively large arctic warming in the
Pre-release Version 24 March, 2005
www.acia.uaf.edu
123
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
ers do not. It is currently not possible to determine which scenario is most likely. In this case, arithmetic averaging of the ACIA-designated model simulations may not be meaningful. It makes more sense to consider all
or at least two scenarios separately: one with and one without a significantly reduced thermohaline circulation
(see also Box 4.5).
• The Barents Sea is currently ice-free throughout the year, mainly as a result of the northward flow of warm
Atlantic water into the region. Models that simulate an ice-covered Barents Sea for the present-day climate
and near ice-free conditions in a climate change scenario will therefore highly overestimate the amplitude of
local warming in the region. In this situation, only model realizations with a realistic sea-ice distribution in the
baseline simulation should be used for a scenario of future climate change.
• Sea ice is generally not well handled by AOGCMs, and this is also the case for the ACIA-designated models.
Specifically, problems relate to their simulation of major polynyas and differences in larger-scale ice cover. For
these areas, future climate change cannot be projected using any specific model or a combination of models.
Given the present state of modeling, expert judgment has to be used in combination with the model
projections.
• Snow cover during winter is reasonably well represented in the Arctic in all of the ACIA-designated models.
However, in both of the transition seasons, as well as during summer, some models exhibit an unrealistically
extensive snow pack for the present-day climate, and then complete absence of snow for some regions in a
climate change scenario. As a consequence, these simulations project temperature increases that are too
large in these regions. In this situation, only model realizations with a realistic seasonal distribution of snow
cover in the baseline simulation should be used for scenarios of future climate change.
Figure 4.16 displays spatial distributions of seasonal temperature changes in the Arctic for the three 21st-century
time slices.The five-model mean projected temperature
2071–2090
2041–2060
2011–2030
light of simulation and projection uncertainties.This
issue is discussed further in section 4.7.
-6
-3
0
+3
Mean temperature change (ºC)
+6
0.0
1.0
1.5
0.5
Standard deviation (ºC)
2.0 0
2
3
4
5
6
1
Mean change/Standard deviation
Fig. 4.14. Changes in annual mean temperature projected by the ACIA-designated models for the early (top row), middle (center
row), and late (bottom row) 21st century, as compared to the ACIA baseline (1981–2000). From left to right: five-model mean
change; the inter-projection standard deviation; and the ratio between the mean and the standard deviation.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
124
Winter
Spring
Summer
Autumn
(Dec–Feb)
(Mar–May)
(Jun–Aug)
(Sep–Nov)
2071–2090
2041–2060
2041–2060
2011–2030
2011–2030
Arctic Climate Impact Assessment
-9
-6
-3
0
+3
Temperature change (ºC)
+6
+9
2071–2090
Fig. 4.16. Seasonal changes in surface air temperature averaged over the five ACIA-designated B2 projections.Top:
2011–2030; middle: 2041-2060; bottom: 2071-2090.
2
4
6
9
12
15
18
Fig. 4.15.The ratio between the ACIA five-model average
projected change in mean annual temperature and √2 times
the standard deviation of 20-year mean temperatures in the
CMIP2 control simulations.
increase over the central Arctic Ocean is greatest in
autumn (up to 9 ºC by 2071–2090), when the air temperature reacts strongly to reduced sea-ice cover and thickness. Average autumn and winter temperatures are projected to rise by 3 to 5 ºC over most arctic land areas. By contrast, projected temperature increases over the Arctic
Ocean in summer remain below 1 ºC, because the temperature is held close to the freezing point by the presence of
melting ice in both the control and the climate change simulations.The contrast between larger temperature increases in autumn and winter and smaller temperature increases
in summer also extends to the surrounding land areas, but
is less pronounced there. In summer, projected temperature increases over northern Eurasia and northern North
America are larger than over the Arctic Ocean, while in
winter the reverse is projected.
The spatial patterns of projected climate change within
the Arctic also differ markedly between the individual
models, so that, at any single location, the scatter of the
model results is larger than it is for change in the arctic
Precipitation change (%)
Fig. 4.17. Projected changes in mean annual surface air temperature in the Arctic (60º–90º N) for the A2 and B2 emissions
scenarios, relative to 1981–2000. A binomial approximation is
used to smooth the original mean of the four ACIA-designated
model projections (CGCM2, ECHAM/OPYC3, GFDL-R30_c, and
HadCM3) for each emissions scenario.
Fig. 4.18. Global and arctic (60º–90º N) percentage change in
mean annual precipitation relative to the baseline period
(1981–2000) projected by the five ACIA-designated models
forced with the B2 emissions scenario.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
125
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Box 4.5.The Atlantic meridional overturning circulation in the 21st century
Evidence from paleoclimate records indicates that the Earth underwent large and rapid climate changes during
the last glacial and early postglacial periods.The origin of the abrupt changes is being closely studied. A number of
studies suggest that the Atlantic meridional overturning circulation (MOC), as part of the global thermohaline circulation, played an active and important role in these rapid climate transitions (e.g., Broecker, 1997; Ganopolski
and Rahmstorf, 2001). Furthermore, modeling studies agree that the Atlantic surface freshwater balance is a key
control parameter for the strength and variability of the Atlantic MOC. A common finding from idealized models
and AOGCMs is that the strength of the Atlantic MOC decreases if the net flux of freshwater to the high northern latitudes increases (Manabe and Stouffer, 1997; Otterå et al., 2003; Rind et al., 2001; Schiller et al., 1997;
Vellinga et al., 2002).
Increased precipitation at high northern latitudes is commonly projected by AOGCM simulations forced with
increasing GHG concentrations (e.g., Räisänen, 2001).There is also recent observational evidence of intensification
of the hydrological cycle at high northern latitudes (Curry R. et al., 2003; Dickson et al., 2002; Peterson et al.,
2002).This raises questions about whether the Atlantic MOC will weaken in the 21st century, and what the likelihood is of a full shutdown of the Atlantic MOC.
Most AOGCMs forced with prescribed scenarios of the major GHG concentrations and aerosol particle distributions project a weakening of the MOC.The projected changes in the maximum strength of the Atlantic MOC by
the end of the 21st century range from about zero to a reduction of 30 to 50% (Cubasch et al., 2001).The projected changes, if any, typically start around 2000 and show a quasi-linear trend thereafter. Irrespective of model
differences in the sensitivity of the simulated Atlantic MOC to global climate change, no AOGCM has yet projected a shutdown of the Atlantic MOC by 2100 (Cubasch et al., 2001).
The present generation of AOGCMs used for these simulations does not include freshwater runoff from melting
ice sheets and glaciers; therefore, it is possible that the model MOC sensitivity to global climate change is too
weak. However, sensitivity experiments with freshwater artificially added to high northern latitudes indicate that
fluxes corresponding to several times the present-day freshwater input are required to significantly alter the
Atlantic MOC (Manabe and Stouffer, 1997; Otterå et al., 2003; Schiller et al., 1997;Vellinga et al., 2002), so this
shortcoming may not be significant. As a result, the IPCC Third Assessment Report (Cubasch et al., 2001) concluded that it is unlikely that the Atlantic MOC will experience a shutdown in the 21st century. It is also likely that
the major part of the North Atlantic–Nordic Seas region will experience warming throughout the 21st century,
even with a weakened Atlantic MOC.
area mean temperature. However, all of the models project substantially smaller temperature increases over the
northern North Atlantic sector than for other parts of the
Arctic. In this area of the sinking branch of the Atlantic
thermohaline circulation, the ocean is well mixed to a
great depth.Therefore, much of the GHG-induced heating is devoted to warming the deeper ocean, rather than
to warming the surface water.The warming is further
reduced because the thermohaline circulation weakens in
most of the model projections, transporting less warm
water from the subtropical regions to the northern
North Atlantic.The different degree of projected weakening of the thermohaline circulation in the models presents a special problem for scenario development for the
North Atlantic area (see Boxes 4.4 and 4.5).
A comparison between changes in mean annual surface air
temperature in the area north of 60º N projected by four
of the ACIA-designated models (CGCM2, ECHAM4/
OPYC3, GFDL-R30_C, and HadCM3) for the A2 and B2
emissions scenarios is shown in Fig. 4.17.The difference
between the two scenarios is not dramatic during the first
half of the 21st century, but becomes more systematic and
significant during the second half of the century.These differences between the projected A2 and B2 arctic areaaveraged temperature increases do not exceed the differences between the highest and lowest projections from the
different ACIA-designated AOGCMs.
4.4.3. Changes in precipitation
Table 4.4. Percentage increases in mean annual precipitation in the Arctic (60º–90º N) compared to the 1981–2000 baseline, as projected by the five ACIA-designated models forced with the B2 emissions scenario (from Kattsov et al., 2003).
Precipitation increase (%)
2011–2030
2041–2060
2071–2090
CGCM2
CSM_1.4
ECHAM4/OPYC3
GFDL-R30_c
2.3
4.6
7.5
8.3
8.3
14.0
5.1
12.3
18.1
3.0
7.4
11.9
Pre-release Version 24 March, 2005
www.acia.uaf.edu
HadCM3
4.0
7.3
12.9
Five-model mean
4.3
7.9
12.3
126
2071–2090
2041–2060
2011–2030
Arctic Climate Impact Assessment
-20
-10
0
+10
Mean precipitation change (%)
+20 0
10
20
Standard deviation (%)
30
-2.0
-1.0
0
+1.0
+2.0
Mean change/Standard deviation
Fig. 4.19. Changes in mean annual precipitation projected by the ACIA-designated models for the early (top), middle (center), and
late (bottom) 21st century, as a percentage of baseline (1981–2000) values. From left to right: five-model mean change; the interprojection standard deviation; and the ratio between the mean and the standard deviation.
Increases in arctic and global mean annual precipitation
over the 21st century projected by the five ACIAdesignated models forced with the B2 emissions scenario
are displayed in Fig. 4.18. By the end of the 21st century
(2071–2090), the projected change in global mean precipitation varies from 1.4% (ECHAM4/OPYC3) to 4.7%
(GFDL-R30_C), with a mean of 2.5%.The Arctic Ocean
and terrestrial arctic regions of North America and
Spring
Summer
Autumn
(Mar–May)
(Jun–Aug)
(Sep–Nov)
2071–2090
2041–2060
2011–2030
Winter
(Dec–Feb)
-30
-20
-10
0
+10
Precipitation change (%)
+20
+30
Fig. 4.20. Seasonal percentage changes in precipitation averaged over the five ACIA-designated B2 projections.Top:
2011–2030; middle: 2041–2060; bottom: 2071–2090.
Eurasia are among the areas with the greatest projected
percentage increase in precipitation.The general increase
in high-latitude precipitation is a robust and qualitatively
well-understood result from climate change experiments.
With increasing temperature, the ability of the atmospheric circulation to transport moisture from lower to
higher latitudes increases, leading to an increase in precipitation in the polar areas where the local evaporation is
relatively small (e.g., Manabe and Wetherald, 1975).
For the area north of 60º N, all five models simulate an
increase in annual precipitation by 2071–2090, which
varies from 7.5% (CGCM2) to 18.1% (ECHAM4/
OPYC3), and from 12 to 14% in the other models
(Table 4.4).The differences between the models increase
rapidly as the spatial domain becomes smaller.
The spatial distribution of the projected mean annual
precipitation changes (five-model average) is shown in
the left column of Fig. 4.19. By 2071–2090, projected
precipitation increases in the Arctic vary from about 5 to
10% in the Atlantic sector to up to 35% locally in the
high Arctic. Most of the projected increase in precipitation is due to increasing atmospheric water vapor convergence, which apparently results from the ability of a
warmer atmosphere to transport more water vapor from
lower to higher latitudes.The increased water vapor convergence is particularly important for precipitation in the
high Arctic, where the local evaporation is small (see section 6.2). Although evaporation is projected to increase
slightly in most of the Arctic, the change is near zero (or
Pre-release Version 24 March, 2005
www.acia.uaf.edu
127
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
even negative) in the Atlantic sector, where the change in
sea-surface temperature is small. Unlike surface air temperature, the ratio between the five-model mean precipitation change and the inter-model standard deviation
(middle column) is larger in the Arctic than almost anywhere else (right column), with comparable agreement
only at high southern latitudes. Nevertheless, the relative
agreement between models is worse for precipitation
than temperature changes.
Similar to projected temperature increases, the projected increase in precipitation is generally greatest in
autumn and winter and smallest in summer (Fig. 4.20),
but the summer minimum is less pronounced than that
of temperature change.
The four-model projected changes in mean annual precipitation for the area north of 60º N (Fig. 4.21) behave
similarly to temperature.The difference between the A2
and B2 scenarios is small in the first half of the 21st century. Later in the century, the difference increases and is
systematic (i.e., shown in all models), but does not
exceed the inter-model scatter.
4.4.4. Changes in other variables
Changes in cloud cover over the Arctic are small but systematic. By 2071–2090, the five-model average projects
an increase in mean annual cloud cover.This is accompanied by a projected decrease (four-model average) in the
mean incident short wave radiation. In summer, this flux
is projected to decrease across the Arctic by more than
10 W/m2 by 2071–2090 compared to the baseline
(1981–2000).
The ACIA-designated models agree in their projections of
decreases in sea-ice and terrestrial snow extents during
the 21st century, as well as general increases both in precipitation minus evaporation over the marine Arctic and
in river discharge to the Arctic Ocean from the surrounding terrestrial watersheds.The projected cryospheric and hydrological changes are quantified and discussed in detail in Chapter 6.
4.4.5. ACIA-designated models in the
CMIP2 exercise
The five ACIA-designated models also participated in the
CMIP2 intercomparison (Meehl et al., 2000), together
with 14 other models.The model versions used in the
CMIP2 simulations are in two cases (CGCM and CSM)
slightly different from the ACIA-designated model versions, but this is unlikely to have any substantial influence on the comparison.The CMIP2 intercomparison
helps to place the ACIA-designated model results in the
broader context of model behavior.The climate changes
in the CMIP2 simulations (Table 4.5) are examined for
the 20-year period centered on the year when atmo-
Winter
Spring
Summer
Autumn
(Dec–Feb)
(Mar–May)
(Jun–Aug)
(Sep–Nov)
2071–2090
2041–2060
2011–2030
The average projected changes in mean seasonal and
annual sea-level pressure are shown in Fig. 4.22.
Throughout the year, there is a slight projected decrease
in pressure in the polar region, suggesting a shift toward
the positive phase of the AO. However, the changes are
small. Even in winter, when the projected decrease in
sea-level pressure over the central Arctic is greatest, this
amounts to only 4 hPa.The changes projected by individual models are larger, but vary widely, especially in
winter.While many impact studies would benefit from
projections of wind characteristics and storm tracks in
the Arctic, the available analyses in the literature are
insufficient to justify any firm conclusions about their
possible changes in the 21st century.
Fig. 4.21. Projected percentage changes in mean annual precipitation in the Arctic (60º–90º N) for the A2 and B2 emissions
scenarios, relative to 1981–2000. A binomial approximation has
been applied to the original mean of the four ACIA-designated
model projections (CGCM2, ECHAM/OPYC3, GFDL-R30_c, and
HadCM3) for each emissions scenario
-4
-3
-2
-1
0
+1
Pressure change (hPa)
+2
+3
Fig. 4.22. Seasonal changes in mean sea-level pressure averaged over the five ACIA-designated B2 projections. For
GFDL-R30_c, the variable used is surface pressure change,
rather than sea-level pressure.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
+4
128
Arctic Climate Impact Assessment
Table 4.5. The 19 models participating in the CMIP2 intercomparison, with the five ACIA models shown in bold.The first column
provides the symbol used in Figure 4.23.
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
aAn
Model acronym
BMRC
CCCmaa
CCSR1
CCSR2
CERFACS
CSIRO
ECHAM3
ECHAM4/OPYC3
GFDL-R15
GFDL-R30_c
GISS
HadCM2
HadCM3
IAP/LASG
LMD/IPSL
MRI1
MRI2
NCAR-CSMa
NCAR/DOE-PCM
Institution
Bureau of Meteorology Research Center
Canadian Centre for Climate Modelling and Analysis
Center for Climate System Research
Center for Climate System Research
Centre Européen de Reserche et de Formation Avancée en Calcul Scientifique
Commonwealth Scientific and Industrial Research Organization
Max-Planck Institute for Meteorology
Max-Planck Institute for Meteorology
Geophysical Fluid Dynamics Laboratory
Geophysical Fluid Dynamics Laboratory
Goddard Institute for Space Studies
Hadley Centre for Climate Prediction and Research
Hadley Centre for Climate Prediction and Research
Institute of Atmospheric Physics, Chinese Academy of Sciences
Laboratoire de Metéorologie Dynamique, Institut Pierre Simon Laplace
Meteorological Research Institute
Meteorological Research Institute
National Center for Atmospheric Research
National Center for Atmospheric Research/ Department of Energy
Reference
Power et al., 1993
Flato et al., 2000
Emori et al., 1999
Nozawa et al., 2000
Barthelet et al., 1998
Hirst et al., 2000
Voss et al., 1998
Roeckner et al., 1999
Manabe et al., 1991
Knutson et al., 1999
Russell and Rind, 1999
Johns et al., 1997
Gordon et al., 2000
Zhang et al., 2000
Braconnot et al., 1997
Tokioka et al., 1995
Yukimoto et al., 2000
Boville and Gent, 1998
Washington et al., 2000
earlier model version than that used by the ACIA participated in the CMIP2 intercomparison.
spheric CO2 doubles in the simulations, which takes 70
years in these idealized experiments.
Figure 4.23a shows the global and arctic (60º–90º N)
mean temperature changes in the CMIP2 simulations.
Models located above the diagonal dashed line project
greater temperature increases in the Arctic than globally,
while models below the dashed line project the reverse.
The global temperature increase varies from 1.1 ºC
(MRI2) to 3.1 ºC (CCSR2), with a mean of 1.75 ºC.The
mean value for the five ACIA-designated models is very
similar but the spread is smaller, 1.4 to 2.0 ºC.The 19model average projected increase in mean annual temperature in the Arctic (60º–90º N) is 3.4 ºC (twice the
global mean), and for the five ACIA-designated models
the projected arctic temperature increase is 3.6 ºC.
Again, the range of the five ACIA-designated model projections (3.1–4.1 ºC) is much smaller than the total
range (1.5–7.6 ºC) of all 19 CMIP2 simulations.
However, individual outliers contribute substantially to
the large range in the CMIP2 results.
Similar conclusions are valid for the projected change in
arctic mean precipitation (Fig. 4.23b).The full range of
projected precipitation change in the 19 models
(4–24%) is wider than the range in the five ACIAdesignated models (8–15%), but the 5-model and the
19-model means are both 11%.The larger set of models
Fig. 4.23. Projections from the 19 CMIP2 models of (a) changes in global mean annual temperature (horizontal axis) and arctic (60º–
90º N) mean annual temperature (vertical axis) and (b) changes in arctic mean annual temperature (horizontal axis) and precipitation
(vertical axis).Table 4.5 lists the model associated with each letter.The five ACIA-designated models are shown in red and the others
in blue.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
129
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
(a) CMIP2,19 Models
(b) CMIP2, 5 Models
Fig. 4.24. Projected increase in mean annual temperature (ºC)
for a doubling of atmospheric CO2 concentration averaged over
(a) the 19 CMIP2 models, and (b) the 5 ACIA-designated models.
shows an expected tendency for the projected precipitation change to be greatest for the models with the greatest projected temperature increase, which is not visible
in the results from the five ACIA-designated models
alone.The same tendency is seen in a comparison of
projected global mean temperature and precipitation
change (see Cubasch et al., 2001).
Figure 4.24 shows the spatial distribution of the projected
change in mean annual temperature from the CMIP2
experiments, as averaged over all 19 models and over the
five ACIA-designated models.The basic patterns are very
similar, with the greatest temperature increases over the
central Arctic in both cases. A similar comparison of projected precipitation changes (Fig. 4.25) also indicates
broad agreement between the 19-model mean and the
5-model mean. However, some of the spatial details in the
5-model mean, such as the maxima over northeastern
Greenland and eastern Siberia, are smoothed out when
the change is averaged over all 19 models.
4.4.6. Summary
Projections of arctic climate change in the 21st century
from all of the ACIA-designated AOGCMs are qualitatively consistent and in line with the IPCC conclusions
listed at the beginning of section 4.4.The across-model
scatter of the arctic climate change scenarios is significant, but smaller than the scatter between the climates
simulated by the different models for the baseline period. Even the difference between the two single-model
projections driven with both the A2 and B2 emissions
scenarios is comparable to the range of corresponding
changes projected by the five ACIA-designated models
forced with the B2 scenario.
In summary, the five ACIA-designated AOGCMs appear
to be a representative sample of climate models, at least
in terms of the average response of arctic temperature
and precipitation to the B2 emissions scenario. However,
this set of projections does not capture the full range of
uncertainty associated with model and emissions scenario
differences, at least in the second half of the 21st century.
(a) CMIP2,19 Models
(b) CMIP2, 5 Models
Fig. 4.25. Projected increase in mean annual precipitation (%) for
a doubling of atmospheric CO2 concentration averaged over (a)
the 19 CMIP2 models, and (b) the 5 ACIA-designated models.
4.5. Regional modeling of the Arctic
An improved understanding of the arctic climate system
is necessary to provide better quantitative assessments of
the magnitude of potential global change and to clarify
the role of the Arctic in the global climate system.The
deficiencies of AOGCMs in describing arctic climate are
partly due to inadequate parameterizations of physical
processes. Equally important, AOGCMs are characterized by a rather coarse horizontal resolution, which fails
to capture atmospheric mesoscale features caused by
coastlines, ice sheets, sea-ice margins, and mountains.To
some extent, this failure is overcome when the resolution is increased (Giorgi et al., 2001).The most obvious
step – to simply increase resolution in AOGCMs – has
until now been impractical due to the computational
capacities required. Experience with high- or variableresolution AGCMs has been limited and no results are
available that focus on the performance of these models
for the Arctic.
Another approach to obtaining enhanced regional details
is the use of a nested limited-area model.This technique
has garnered considerable interest in recent years
because it requires less computational capacity than a
global model with comparable resolution.These models
can be used for process studies and for model validation
studies using lateral boundaries from observation-based
analysis (e.g., Giorgi and Mearns, 1999; Giorgi et al.,
2001). However, the technique can also be used as the
basis for dynamic downscaling of AOGCM simulations.
Models used in this way are often referred to as regional
climate models (RCMs; see Box 4.1).
4.5.1. Regional climate models of the arctic
atmosphere
4.5.1.1. General
Regional models show considerable skill in short-term
(few hour to few day) weather forecasting and are used
worldwide for this purpose.This application is basically
an initial-value problem, and most of the success of this
approach can be ascribed to the high resolution of the
models and the use of very recent observations.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
130
Arctic Climate Impact Assessment
Likewise, at timescales of a few years to decades and
beyond, regional models have shown their strength in
comparison with coarse-resolution global models, as they
are capable of capturing the fine-scale details of climatic
processes – such as the presence of complex topography
and small-scale weather features such as tropical cyclones
and even polar lows – much more realistically than global
models (Giorgi et al., 2001). Moreover, it has been
shown that the statistics of extreme precipitation are realistically simulated by high-resolution regional models
(Frei et al., 2003b) and that they show skill even at their
grid scale (Huntingford et al., 2003).This is a boundaryvalue problem and the assessment of the skill of the
model is based on statistics of model performance over
long time periods (several years or more).
Regional model projections are limited by the quality of
the global model projections used for the lateral boundary conditions. In this respect, a potential problem is the
mismatch between scales in the coarse-resolution global
model and the high-resolution regional model. Recently,
this has been demonstrated not to be a fundamental
problem if proper boundary condition procedures are
used (Denis et al., 2002). In fact, it has been demonstrated that, when driven by realistic (observed) boundary conditions, regional climate models are capable of
capturing the overall observed regional climatic evolution and can add realistic spatial and temporal information to the information provided by the driving model
(Dethloff et al., 2002; Frei et al., 2003b; Giorgi et al.,
2001).When nested within a GCM, it is important,
however, to stress that the large scale circulation is
imposed by the lateral boundaries.The regional model is
not able to, nor is it intended to, correct the large-scale
errors made by the global model.The role of the regional model is instead to add regional detail and fine spatial
and temporal scales to the simulation, not to improve
the large-scale simulation.
Atmospheric regional modeling systems for the Arctic
have been developed by Lynch et al. (1995) and Dethloff
et al. (1996). Dethloff et al. (1996) applied the regional
atmospheric climate model HIRHAM (Christensen J. et
al., 1996) to the entire Arctic north of 65º N.This
model has so far been the only RCM used with a circumpolar focus for climate simulations, although the
recent Arctic Regional Model Intercomparison Project
(ARCMIP; see section 4.5.1.2) initiative has increased
the interest in arctic RCMs. Many other groups have
developed RCMs, but mostly with a more southerly
region of interest, although in some cases parts of the
Arctic have been included in simulations performed with
these models. For example, several models have been
applied over most of the European continent including
the Scandinavian Peninsula and parts of the North
Atlantic Ocean extending all the way to the ice margins;
even parts of Greenland have been included in such simulations (Christensen J. et al., 1997; Rummukainen et
al., 2001). Likewise, experiments with a Canadian RCM
have been applied to the whole of Canada (Laprise et
al., 1998) including a fair proportion of the Arctic.The
development of these models continues, but limited
results with an arctic focus have been published.
Giorgi et al. (2001) documented how much insight has
been provided into fundamental issues concerning the
nested regional modeling technique. For example,
multi-year to multi-decadal simulations must be used for
climate change studies to provide meaningful climate
statistics, to identify significant systematic model errors
and climate changes relative to internal model and
observed climate variability, and to allow the atmospheric model to equilibrate with the land surface conditions (e.g., Christensen O., 1999; Jones et al., 1997;
Machenhauer et al., 1998; McGregor et al., 1999). In
addition, the choice of domain size is not a trivial question.The influence of the boundary forcing is reduced as
region size increases (Jacob and Podzun, 1997; Jones et
al., 1995) and may be dominated by the internal model
physics for certain variables and seasons (Noguer et al.,
1998).This can lead to the RCM solution significantly
departing from the driving data, which can make the
interpretation of downscaled regional climate changes
more difficult (Jones et al., 1997). For most experiments with very high resolution, the domain size is limited by practical considerations and the large-scale flow
is, therefore, constrained substantially by the driving
model (Christensen J. and Kuhry, 2000). Denis et al.
(2002) demonstrated that when the discrepancy in resolution between the model providing the lateral boundary
conditions and the nested RCM does not exceed a factor
of 10, the RCM is able to generate added value highresolution information.With this limitation, they also
showed that for a typically sized domain, the RCM does
not introduce any spurious developments due to the
nesting technique, and is fully capable of a consistent
development within the model domain.
Configurations of RCM model physics are derived either
from a pre-existing (and well-tested) limited-area model
system with modifications suitable for climate applications
or are implemented directly from a GCM (see Giorgi et
al., 2001). In the first approach, each set of parameterizations is developed and optimized for the respective model
resolutions. However, this makes interpreting differences
between the nested model and the driving GCM more
difficult, as these will result from more than just changes
in resolution. Also, the different model physics schemes
may result in inconsistencies near the boundaries
(Machenhauer et al., 1998; Rummukainen et al., 2001).
The second approach maximizes compatibility between
the models. However, physics schemes developed for
coarse-resolution GCMs may not always be adequate for
the high resolutions used in nested regional models and
may at a minimum require recalibration. Overall, both
strategies have shown performance of similar quality and
either is acceptable (Giorgi and Mearns, 1999).
4.5.1.2. Simulations of present-day climate
with regional climate models
Pre-release Version 24 March, 2005
www.acia.uaf.edu
131
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Regional climate models have been used for a wide variety of research worldwide, and have generated a sizeable
research community. However, very few groups have
focused on the Arctic to date, although several new initiatives have recently been undertaken. Intercomparison
projects entailing participation of different international
research groups have been conducted or are currently
underway (Christensen J. et al. 1997, 2002;
Machenhauer et al., 1998;Takle et al., 1999; Rinke et
al., 2000; www:awi-potsdam.de/www-pot/atmo/
glimpse/index.html).This section describes the existing
applications of RCMs for simulating present-day arctic
climate conditions.
Table 4.6. Comparison of observed changes in net accumulation rates for the north Greenland Ice Sheet (76º–79º N) with
simulations by a general circulation model (ECHAM4/OPYC3)
and a regional climate model (HIRHAM). Model uncertainties
are estimated as one standard deviation of the 30-year interannual variability in the simulation.
Advances in regional climate modeling must be based on
an analysis of physical processes and comparison with
observations. In data-poor regions such as the Arctic, this
procedure may be complemented by a community-based
approach (i.e., through collaborative analysis by several
research groups).To illustrate this approach, simulations
of the Arctic Basin north of 65º N have been performed
and planned with different RCMs, driven at their lateral
boundaries by observationally based analyses (Rinke et
al., 2000). Motivated by this, an international intercomparison of regional model simulations for the Arctic,
ARCMIP, has been organized under the auspices of the
World Climate Research Programme.The foci of the
evaluation include the surface energy balance over ocean
and land, clouds and precipitation processes, stable planetary boundary layer turbulence, ice-albedo feedback, and
sea-ice processes.The preliminary results from this project indicate that the participating regional models are
able to reproduce reasonably well the main features of
the large-scale flow and the surface parameters in the
Arctic. However, in order to reach definitive conclusions
in an RCM intercomparison, ensemble simulations with
adequate spin-up (time for regional model processes to
equilibrate with prescribed external forcings) and equivalent initialization of surface fields are required (Rinke et
al., 2000). Several aspects of the intercomparison are difficult due to the lack of adequate data for model validation. However, some aspects of the models, such as the
hydrological cycle, compare well with each other (see
Rinke et al., 2000).
Change in net accumulation rate (mm/yr)
ECHAM4/OPYC3a
HIRHAMa
Observedb
a1961–1990
b1954–1995
27º–30º W
170±40
75±24
97±84
60º–65º W
-1390±560
-223±189
-310±107
(Kiilsholm et al., 2003)
(Paterson and Reeh, 2001)
(a)
Determining the primary causes of model biases, deficiencies, and uncertainties in atmosphere-only and coupled atmosphere-ice-ocean climate models for the Arctic
is of vital importance in order to model the arctic climate adequately. Participants in the ARCMIP project are
seeking to improve model representations through an
intercomparison of simulations by different models and
comparison with observations made during the Surface
Heat Budget of the Arctic Ocean field-experiment year
(October 1997 to October 1998).
(b)
(kg/yr)
1000
600
400
200
100
25
10
5
0
-5
-10
-25
-100
-200
-400
-600
Fig. 4.26. Net mass balance of the Greenland Ice Sheet over
the period 1961–1990 simulated by (a) the ECHAM4/OPYC3
general circulation model and (b) the HIRHAM regional climate model. Model-designated elevation contours shown for
every 500 m (Kiilsholm et al., 2003).
For the Arctic, only very limited multi-year RCM experiments have been conducted. For example, Kiilsholm et
al. (2003) assessed the uncertainty in regional accumulation rates for the Greenland Ice Sheet due to model resolution.They used the HIRHAM RCM (50 km resolution) with boundary conditions from a 30-year control
simulation with the ECHAM4/OPYC3 model (~300 km
horizontal resolution; Roeckner et al., 1996). Figure
4.26 compares the resulting accumulation and ablation
zones as simulated by the RCM and the GCM.Table 4.6
illustrates the ability of the models to simulate present
regional (northern Greenland) changes in accumulation
rates. It appears that the RCM simulation is in better
agreement with the observational evidence than the
GCM simulation.
Non-arctic applications have shown that regional models
may have significant potential for use as dynamic interpolators, yielding useful data for a wide range of times and
Pre-release Version 24 March, 2005
www.acia.uaf.edu
132
Arctic Climate Impact Assessment
(a)
4.5.1.3.Time-slice projections from
atmospheric RCMs
(b)
Accumulation rates (cm/yr)
>3000
1500–3000
1000–1500
900–1000
800–900
700–800
600–700
500–600
400–500
300–400
200–300
180–200
160–180
140–160
120–140
100–120
90–100
80–90
<80
Fig. 4.27. Net accumulation rates over the Greenland Ice
Sheet (a) estimated from observations and (b) four years of
HIRHAM simulation (from Dethloff et al., 2002).
locations where in situ observations are not available. As
regional models become increasingly accurate, they could
become valuable tools for glaciological research. A primary application in glaciology is the investigation of mass
balance changes in continental ice sheets. Determining
the climatic conditions affecting ice sheets is important
because major changes in ice-sheet dimensions affect climate and sea level throughout the world. Much recent
work has gone into the validation of RCM simulations
(Bromwich et al., 2001; Cassano et al., 2001) using
observational data analyses (Hanna and Valdes, 2001) for
the Greenland Ice Sheet. Dethloff et al. (2002) carried
out a detailed RCM validation study on the basis of
multi-year ensemble simulations, selected annual simulations, and derived results for the mass balance of the
Greenland Ice Sheet. Figure 4.27 illustrates net accumulation rates from the HIRHAM simulations. Compared
with available results from earlier work, these results
indicate a high degree of skill in spatial representation.
Another promising application was identified by
Christensen J. and Kuhry (2000), who analyzed the ability of an RCM to simulate permafrost zonation at very
high spatial resolution. Based on a simple permafrost
index (Nelson and Outcalt, 1987), but applied to the
subsurface model layers rather than near-surface air temperatures, they documented that at sufficiently high resolution the permafrost zonation in complex regions can
be quite accurately modeled.
In IPCC (2001), the concept of using RCMs for climate
change projections was considered at some length.The
IPCC (2001) considered the concept of using RCMs for
climate change projections at some length, and concluded that it is essential that information from a transient
AOGCM simulation be available. In such studies, the
RCM is used to provide a reinterpretation of the overall
AOGCM behavior, including its response to external
forcing (from GHGs and aerosols). Sometimes ocean
modeling is also done as part of the regional model simulation (Räisänen et al., 2003). In a typical experiment
(Christensen J. et al., 2002; Jones et al., 1995; Kiilsholm
et al., 2003; Machenhauer et al., 1996; Räisänen et al.,
1999; Rummukainen et al., 2001), two time slices (e.g.,
1961–1990 and 2071–2100) are selected from a transient AOGCM simulation.The RCM simulations include
prescribed time-dependent GHG and aerosol concentrations for the corresponding periods of the AOGCM run.
The time-dependent sea-surface temperatures and seaice distributions simulated by the AOGCM are also prescribed as lower boundary conditions, although some
models also incorporate an interactive ocean/sea-ice
model (Räisänen et al., 2003).The RCM simulations are
typically initialized using atmospheric and land-surface
conditions interpolated from the corresponding
AOGCM fields, and there may be a considerable spin-up
period before the actual simulation is started (e.g.,
Christensen O., 1999).
Only a few studies of this type have been conducted for
time slices of durations long enough to encompass the
large interannual variability in Arctic. Only two sets of
experiments exist to date that cover the entire Arctic
(Dorn et al., 2003; Kiilsholm et al., 2003), while
Haugen et al. (2000) report simulations that only cover
the Atlantic sector, with the main focus of the experiment being Norwegian land territories (mainland
Norway and Svalbard). Kiilsholm et al. (2003) and Dorn
et al. (2003) have conducted a set of such experiments
using information from transient simulations with the
ECHAM4/OPYC3 model (Stendel et al., 2000).
This section highlights the experiments conducted by
Kiilsholm et al. (2003) using the HIRHAM4 model
forced with the B2 emissions scenario, as they are the
only ones with complete multi-year integrations that are
generally consistent with the characteristics of the ACIA
scenarios. In this study, the AOGCM had a resolution of
approximately 300 km, while the resolution of the RCM
was approximately 50 km.This discussion focuses on differences in the climate projections of the RCM and the
driving AOGCM.
Figure 4.28 depicts the winter temperature change as
simulated by the AOGCM and the RCM. In general, the
patterns of warming as well as their amplitude are quite
similar. However, the RCM depicts some regional patterns
that can be ascribed to the higher resolution. Along the ice
Pre-release Version 24 March, 2005
www.acia.uaf.edu
133
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
(a) ECHAM4/OPYC3
(b) HIRHAM4
(ºC)
10
8
6
5
4
3
Fig. 4.28. Winter (Dec–Feb) temperature increase in the
Arctic between 1961–1990 and 2070–2100 projected by (a)
the ECHAM/OPYC3 general circulation model and (b) the
HIRHAM4 regional climate model (modified from Rysgaard
et al., 2003).
margin in the Greenland Sea and the Barents Sea, where
sea ice retreats in the simulations (sea ice in the RCM is
interpolated from the AOGCM), the RCM shows greater
temperature increases.This is due to a stronger response
to sea-ice changes resulting from a better description of
the nonlinear energy cascade connected with mesoscale
weather developments (e.g., stronger cyclonic developments) in the RCM than in the AOGCM. Conversely,
temperature increases projected by the RCM tend to be
lower over most of the central Arctic and all of Siberia,
particularly during summer.This is due to a more realistic
simulation of the present-day snow pack by the RCM than
by the AOGCM (see Box 4.4).
Figure 4.29 shows the winter change in simulated precipitation minus evaporation. As with temperature, the
large-scale agreement is striking. However, regional
details are evident along the North Atlantic storm track
and close to complex topography. In general, the RCM
shows a stronger increase in precipitation minus evaporation upwind of major topographical obstacles and a
corresponding decrease in precipitation minus evapotranspiration downwind (e.g., Scandinavia, the Rocky
Mountains, and Siberia).This is partly explained by the
increased topographical gradients due to the higher resolution of the model.
These apparently small differences may have substantial
effects on the assessment of future changes in various
geophysical systems, such as the mass balance of glaciers
and the Greenland Ice Sheet in particular (Kiilsholm et
(a) ECHAM4/OPYC3
(b) HIRHAM4
(mm/d)
al., 2003) and the depth of the permafrost active layer
(Walsh, in press; see also section 6.6).The RCM is better suited for simulating climate change at the regional
level, particularly for areas with complex topography
and coastlines.This has been confirmed in multiple
applications of the model outside of the Arctic (e.g.,
Giorgi et al., 2001; Christensen J. and Christensen O.,
2003; Huntingford et al., 2003).
Projected changes in arctic climate due to anthropogenic
GHG emissions will occur together with natural dynamic processes in the climate system. In order to improve
projections of the evolution of arctic climate, the effects
of natural climate variability and in particular their
regional dimensions must be taken into account. One
major phenomenon contributing to the natural variability of the climate of the Northern Hemisphere is the
NAO, which is also associated with the AO, and is
described in more detail in section 2.2.2. In general, the
influence of the NAO on arctic temperatures is directly
opposed in the western and eastern Arctic, and is
stronger over land areas than over ocean areas or sea ice.
Dorn et al. (2003) investigated the combined effects of
varying phases of the NAO and increasing GHG and
aerosol concentrations on arctic winter temperatures
(Fig. 4.30). In this study, different phases of the NAO in
a transient coupled AOGCM simulation were considered
in time slice simulations using an RCM.Two future periods, 2013–2020 and 2039–2046, representing a positive
and a negative phase of the NAO, respectively, were analyzed.The level of GHGs and aerosols in the simulation
is higher during the negative NAO phase (2039–2046)
than during the positive phase (2013–2020). Although
mean arctic winter temperatures are projected to be
approximately 1.3 ºC higher during the negative phase
compared to the positive phase, regions of warming and
cooling between the two periods can be observed in
Figs. 4.30a and 4.30b. Subsequent to the positive phase,
a strong warming of more than 6 ºC is simulated over
some areas of Alaska, the Labrador Sea, and Baffin
Island, whereas a similar strong cooling is simulated only
over the northern Barents Sea.The temperature effect of
the NAO is altered by the general temperature increase
resulting from enhanced GHG and aerosol concentrations, but the influence of the NAO is still clearly evident at the regional scale. Although the statistical signifi(a) 2013–2020
(b) 2039–2046
(ºC)
>12
1.0
10–12
0.5
8–10
0.2
6–8
0.1
4–6
-0.1
-0.2
2–4
-0.5
0–2
<0
Fig. 4.29. Winter (Dec–Feb) change in precipitation minus
evaporation between 1961–1990 and 2070–2100 projected
by (a) the ECHAM/OPYC3 general circulation model and (b)
the HIRHAM4 regional climate model (modified from
Rysgaard et al., 2003).
Fig. 4.30. Projected change in monthly mean winter (Dec–Mar)
temperature at 2 m height compared to the control climate for
(a) 2013–2020, when the simulated North Atlantic Oscillation
(NAO) was in a positive phase; and (b) 2039–2046, when the
simulated NAO was in a negative phase (Dorn et al., 2003).
Pre-release Version 24 March, 2005
www.acia.uaf.edu
134
Arctic Climate Impact Assessment
cance of these differences has not been assessed, the
results clearly show that regional changes in the Arctic at
decadal timescales may, at least for several more
decades, be dominated by changes in the overall atmospheric flow rather than by the temperature increases
due to rising GHG concentrations.
Circulation in an RCM is determined by the boundary
conditions provided by the driving AOGCM. As noted in
section 4.3, simulations from different global models can
be very different in terms of circulation patterns, and
the small-scale response in a regional model can amplify
such differences. An example is given in Fig. 4.31. Here,
the Rossby Centre regional model (RCAO:
Rummukainen et al., 2001) is driven by 30-year global
climate simulations from the HadAM3H and
ECHAM4/OPYC3 global climate models forced with
the A2 and B2 emissions scenarios. Both regional simulations driven by the HadAM3H scenarios (Fig. 4.31a,b)
show only moderate increases in precipitation while
both ECHAM4/OPYC3 simulations (Fig. 4.31c,d) show
dramatic precipitation increases, particularly on the
western side of the Scandinavian mountains.The differ(a)
(b)
ence between the A2 and B2 emissions scenarios is quite
small relative to the inter-model differences, which are
due to a clear difference in the circulation regime
change simulated by the two AOGCMs.The HadAM3H
model projects a relatively small change in the northsouth pressure gradient across the Nordic region, in
sharp contrast to the ECHAM4/OPYC3 model, which
projects a substantial strengthening of the north-south
pressure gradient.This difference is in turn connected
with a difference in the projected shifts in the main
storm track regions over the North Atlantic.The physical reason behind the different responses in storm track
regions for the two global models is at present unclear.
Keeping in mind the results of Dorn et al. (2003), a
plausible explanation is that part of the difference may
be due to different simulations of decadal variability in
the NAO by the two AOGCMs.
The lack of experiments with different RCMs using
boundary conditions from more than one GCM so far
prevents the practical use of RCM information for general impacts work.This is also the situation even for
better-studied regions, such as continental Europe, due
to a lack of appropriate coordination of these efforts;
however, see Christensen J. et al. (2002) for an update
on recent progress.
4.5.2. Regional Arctic Ocean models
(c)
-30
Proper treatment of the Arctic Ocean requires a spatial
resolution high enough to account for reduced Rossby
length scales (the smallest scale at which the rotation of
the Earth has a dominating influence on flow dynamics);
to permit the important flows through Bering Strait and
the Canadian Archipelago; and to accurately represent
the complex bottom topography steering the currents,
as well as the continental slopes and the large continental shelves where the thermohaline, wind-driven, and
tidal dynamics interact.
(d)
-20
-10
20
30
10
0
Precipitation change (%)
40
50
60
Fig. 4.31. Percentage changes in winter (Dec–Feb) precipitation over the Scandinavian region between 1961–1990 and
2071–2100 as simulated by a regional climate model (RCAO)
driven by (a) the HadAM3H model forced with the A2 emissions scenario; (b) the HadAM3H model forced with the B2
emissions scenario; (c) the ECHAM4/OPYC3 model forced
with the A2 emissions scenario; and (d) the ECHAM4/OPYC3
model forced with the B2 emissions scenario.
Currently, about a dozen Arctic Ocean models exist,
most of which are participating in the Arctic Ocean
Model Intercomparison Project (AOMIP: Proshutinsky
et al., 2001; see also Box 4.2). Most of the Arctic Ocean
models are derived from global oceanic GCMs – in
many cases from the Geophysical Fluid Dynamics
Laboratory (GFDL) Modular Ocean Model (MOM:
Pacanowski and Griffies, 1999).The Arctic Ocean models represent a wide spectrum of numerical approaches,
employing either finite-difference (in most cases) or
finite-element approximations, and three types of vertical coordinates: z (constant geopotential surfaces), sigma
(bathymetry-following), and isopycnal (constant potential density referenced to a given pressure).The models
differ in their spatial resolution (from 40–50 km down
to 16–20 km in the horizontal and typically 25–30 levels
in the vertical); specifications of surface and lateral fluxes; and formulations of the surface mixed and bottom
boundary layers. Regional models of the Arctic Ocean
are usually coupled to comprehensive sea-ice models,
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
135
although these often differ in their treatment of dynamics and thermodynamics.
model simulations as it does in observations inferred
from satellite data.
To ensure a degree of fidelity to the simulations, an artificial constraint known as “climate restoring” (i.e., relaxation to the observed climate) based on the surface salinity is often introduced.This constraint prevents a model
from drifting significantly from observations. Restoring
time constants vary across models from several months
(strong restoring) to several years (weak restoring).
Some models do not employ restoring at all. However,
the biases in simulations monotonically increase with the
value of the restoring constant, reaching their highest
levels in the models without restoring (Proshutinsky et
al., 2001).This emphasizes that there are some processes
and feedbacks crucial for representing the Arctic Ocean
general circulation that are still neither sufficiently
understood nor properly represented in the regional
models, just as is the case for the global models.
The HIRHAM-MOM coupled model (Rinke et al.,
2003) reproduced the general sea-level pressure patterns
for the summer of 1990, based on a comparison with
ECMWF analyses. However, discrepancies appeared in
late summer that significantly affected variables such as
wind flow and sea-ice transport. Similar to the ARCSyM
results (Maslanik et al., 2000), HIRHAM-MOM simulated too much sea ice in the Bering, Chukchi, and East
Siberian Seas during the summer of 1990.This is also the
case for the ocean/sea-ice models driven by ECMWF
atmospheric data.
From the viewpoint of their employment in downscaling
AOGCM outputs or in constructing scenarios of future
climate change, regional ocean models lag behind the
atmospheric RCMs.
4.5.3. Coupled arctic regional climate models
The construction of coupled RCMs is a recent development.These models couple atmospheric RCMs to other
models of climate system components, such as lake,
ocean/sea-ice, chemistry/aerosol, and land biosphere/
hydrology models (Bailey and Lynch, 2000a,b; Bailey et
al., 1997; Hostetler et al., 1994; Lynch et al., 1995,
1997, 1998; Mabuchi et al., 2000; Maslanik et al., 2000;
Qian and Giorgi, 1999; Rinke et al., 2003; Roed et al.,
2000; Rummukainen et al., 2001; Small et al., 1999a,b;
Tsvetsinskaya et al., 2000;Weisse et al., 2000).These initial efforts provide a path toward the development of coupled “regional climate system models”. For some parts of
the Arctic, coupled mesoscale atmosphere-ice-ocean
models already exist, although they are restricted to small
domains and short integration times (e.g., Lynch et al.,
1997, 2001; Roed et al., 2000; Schrum et al., 2001).
For the circumpolar Arctic, Maslanik et al. (2000) presented the first results of a coupled atmosphere-iceocean RCM called ARCSyM.The oceanic component in
this RCM was a simple mixed layer model. Rinke et al.
(2003) present a more complex coupled RCM (i.e., a
fully coupled atmosphere-ice-ocean circulation model
system) called HIRHAM-MOM.The ability to simulate
conditions over the Arctic Ocean during April to
September 1990, a period of anomalous atmospheric
circulation and sea-ice conditions, was investigated with
both models. A common result was found: neither
model was able to correctly reproduce the large retreat
of sea ice in the eastern Eurasian Basin and the adjacent
shelf sea observed during the summer of 1990 (Maslanik
et al., 1996).The sea ice in the Chukchi and East
Siberian Seas does not retreat as completely in the
The results from both coupled models highlight the
importance of regional atmospheric circulation in driving interannual variations in arctic sea-ice extent, and
illustrate the level of model performance required to
simulate such variations. Such studies are valuable
because they indicate improvements needed in the models by evaluating the results against observations, and the
roles of key processes and feedbacks by comparing the
results to those of the uncoupled atmospheric model.
While results for the Baltic Sea imply improved model
performance when an ocean model is coupled to the
RCM (Räisänen et al., 2003), high-resolution coupled
model systems for the Arctic have not provided
improved performance to date.
4.5.4. Summary
The current status of arctic regional climate modeling
did not allow RCMs to be employed as principal tools
for the ACIA. Present scenarios of future arctic climate
change are therefore based on results from global
AOGCMs. However, presently available global coupled
models have a coarse spatial resolution that limits their
ability to capture many important aspects of climate
change. In particular, intense storms, the effects of
topography, and fundamental aspects of regional ocean
circulation cannot be represented adequately.To improve
the modeling of such phenomena, development of
regional coupled ocean-ice-atmosphere climate models
should receive a high priority in the efforts of the climate modeling community. Such developments should
go hand in hand with developments in global modeling,
in particular development of high-resolution AOGCMs.
4.6. Statistical downscaling approach and
downscaling of AOGCM climate change
projections
Statistical downscaling (also called empirical downscaling) is a tool for downscaling climate information from
coarse spatial scales to finer scales. It may be applied as
an alternative, or as a supplement, to dynamic downscaling (i.e., regional modeling).The underlying concept is
that local climate is conditioned by large-scale climate
and by local physiographical features such as topography,
distance to a coast, and vegetation (von Storch, 1999). At
Pre-release Version 24 March, 2005
www.acia.uaf.edu
136
Arctic Climate Impact Assessment
a specific location, therefore, links should exist between
large-scale and local climatic conditions. Statistical downscaling consists of identifying empirical links between
large-scale patterns of climate elements (predictors) and
local climate (predictand), and applying them to output
from global or regional models. Successful statistical
downscaling is thus dependent on long reliable series of
predictors and predictands. Giorgi et al. (2001) provide a
survey of statistical downscaling studies with emphasis on
studies published between 1995 and 2000.
4.6.1. Approach
4.6.1.1. Predictands
Although mean temperature and precipitation (seasonal,
monthly, or daily) are the most commonly used local predictands, statistical downscaling has also been applied to
generate local scenarios of cloud cover, daily temperature
range, extreme temperatures, relative humidity, sunshine
duration, snow-cover duration, and sea-level anomalies
(Enke and Spekat, 1997; Heyden et al., 1996; Kaas and
Frich, 1995; Martin et al., 1997; Schubert, 1998; Solman
and Nuñez, 1999). Even sea ice (Omstedt and Chen,
2001), ocean salinity and oxygen concentrations (Zorita
and Laine, 2000), zooplankton (Heyden et al., 1998), and
phytoplankton spring bloom in a Swedish lake (Blenckner
and Chen, 2003) have been used as predictands.
4.6.1.2. Predictors
The large-scale predictors should satisfy certain conditions: they should be reproduced realistically by the particular global model; they should (alone or combined)
be able to account for most of the observed variations in
the predictand; the statistical relationships should be
physically interpretable and temporally stationary; and,
when applied to a changing climate, predictors that
“carry the climate change signal” should be included
(Giorgi et al., 2001).
The optimal choice of predictors depends upon the predictand. For downscaling local temperature, large-scale
fields of geopotential height or air temperature might be
used as the “signal-carrying” predictors (e.g., Huth,
1999). For precipitation, absolute or specific humidity
may be used (Crane and Hewitson, 1998; Hellström et
al., 2001;Wilby and Wigley, 2000). In maritime regions,
air temperature can sometimes serve as a proxy for
humidity (Wilby and Wigley, 2000). In addition, some
indicator of atmospheric circulation (e.g., sea-level pressure or a geopotential height field) is usually included
(e.g., Chen and Chen, 2003).
4.6.1.3. Methods
Surveys of methods for establishing links between largescale predictors and local predictands are provided by
Hewitson and Crane (1996), Zorita and von Storch
(1997),Wilby and Wigley (1997), Xu (1999), Giorgi et
al. (2001), and Mearns et al. (2001).The choice of
method should depend on predictand, time resolution,
and also on the application of the scenario. Linear methods such as canonical correlation analysis (CCA), singular
value decomposition (SVD), and multiple linear regression analysis (MLR) can, in most cases, be used to generate scenarios of monthly or seasonal values (e.g., Busuioc
et al., 1999; Corte-Real et al., 1995; Huth and Kysely,
2000; Sailor and Li, 1999).To generate scenarios for variables such as daily precipitation, however, nonlinear techniques such as weather classification (e.g., Conway and
Jones, 1998; Enke and Spekat, 1997; Goodess and
Palutikof, 1998, Palutikof et al., 2002; Schnur and
Lettenmaier, 1997), neural nets (Cavazoz, 1999; Clair and
Ehrman, 1998; Crane and Hewitson, 1998; Schoof and
Pryor, 2001), or analogues (Zorita and von Storch, 1999)
are most useful.Weather generators (e.g., Semenov and
Barrow, 1997; Semenov et al., 1998;Wilby et al., 1998;
Wilks, 1999) can also be applied for generating scenarios
with daily resolution, starting from monthly climatechange scenarios generated by one of the above methods.
4.6.1.4. Comparison of statistical downscaling
and regional modeling
Several studies have compared results from statistical and
regional modeling (Cubasch et al., 1996; Hellström et
al., 2001; Kidson and Thompson, 1998; Murphy, 1999,
2000).The main impression from these studies is that
results from the two downscaling techniques are usually
quite similar for present-day climate, while differences
in future climate projections are found more frequently.
These differences can, to a large degree, be explained by
the unwise choice of predictors in the statistical downscaling, for example, predictors that carry the climate
signal (Murphy, 2000). It has also been suggested that
results from statistical downscaling may be misleading
because the projected climate change exceeds the range
of data used to develop the model (Mearns et al., 2001).
However, differences between results from statistical
downscaling and regional modeling may also result from
the ability of statistical downscaling to reproduce local
features that are not resolved in the regional models
(Hanssen-Bauer et al., 2003).
Some disadvantages of statistical downscaling versus
regional modeling are as follows.
• The major weakness of statistical downscaling is the
assumption that observed links between large-scale
predictors and local predictands will persist in a
changed climate.
• A problem when applying statistical downscaling
techniques to daily values is that the observed
autocorrelation between the weather at consecutive time steps is not necessarily reproduced. If it
is essential to reproduce this, a suitable method
(e.g., weather generators; Katz and Parlange,
1996;Wilks, 1999) should be used.
• Statistical downscaling does not necessarily reproduce a physically sound relationship between different climate elements. Using a downscaling
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
method based on weather classification for several
predictands (e.g., Enke and Spekat, 1997) can minimize this problem.
• Successful statistical downscaling depends on long,
reliable observational series of predictors and predictands.
Some advantages of statistical downscaling versus regional modeling are as follows.
• Statistical downscaling is less technically demanding than regional modeling. It is thus possible to
downscale from several GCMs and several different emissions scenarios relatively quickly and inexpensively (Benestad, 2002).
• It is possible to tailor scenarios for specific localities, scales, and problems.The spatial resolution
applied in regional climate modeling is still too
coarse for many impact studies, and some variables
are either not available or not realistically reproduced by regional models. For example, Omstedt
and Chen (2001) applied statistical downscaling to
infer sea-ice extent in the Baltic Sea.
• In most cases, the development of statistical downscaling models includes an evaluation of AOGCM
performance in simulating the climate of a specific
region (Busuioc et al., 1999, 2001a). Methods
applied in statistical downscaling have been used to
evaluate large-scale fields of single variables as well
as the links between different fields (Busuioc et al.,
2001b; Hanssen-Bauer and Førland, 2001;Wilby
and Wigley, 2000).The stability of these links under
global change has also been investigated (e.g., Chen
and Hellström, 1999). Such analyses can indicate
which variables serve as the best predictors.
4.6.2. Statistical downscaling of AOGCM
climate change projections in the Arctic
A survey of statistical downscaling studies up to 2000 is
provided by the IPCC (Giorgi et al., 2001). However,
very few of these studies considered arctic sites. During
the 1990s, a few statistical downscaling studies in the
Atlantic/European sector of the Arctic were performed
at the Danish Meteorological Institute. Kaas and Frich
(1995) downscaled monthly means of diurnal temperature range (DTR) and cloud cover at ten synoptic stations from Greenland in the west to Finland in the east.
The 500 hPa height and the 500/1000 hPa thickness
anomaly fields were used as predictors in an MLR-based
model.The model predictors were taken from the final
30 years of a control simulation of the 20th century and
the final 30 years of a scenario “A” (business as usual)
simulation of the 21st century generated by the
ECHAM1 model (Cubasch et al., 1992). Kaas and Frich
(1995) found that statistically significant negative trends
in DTR were projected for Fennoscandia, especially in
central and eastern areas, and especially during winter.
For Greenland and Iceland, only minor trends in DTR
were projected. Positive trends in cloud cover were pro-
137
jected over most of the area; these were most significant
in northeastern areas of Fennoscandia.
In Canada, artificial neural networks (ANNs) have been
applied to model hydrological variables (Clair and
Ehrman, 1998; Clair et al., 1998; Ehrman et al., 2000).
Clair et al. (1998) present a scenario for changed runoff
from different Canadian ecozones (including arctic
areas) under conditions of doubled atmospheric CO2
concentrations. A doubled-CO2 equilibrium climate
change scenario produced by the Canadian climate
model CCC (Boer et al., 1992) was used to generate
scenarios for individual basins.Temperature and precipitation scenarios were fed into the ANN to produce scenarios of changes in runoff. In the arctic ecozones that
were investigated in the study, the projected changes in
annual runoff were between 0 and +10%, the spring
melt advanced by between a couple of weeks and one
month, and there was a tendency for reduced runoff
during summer. Qualitatively similar findings are reported in section 6.8, where annual discharge from various
North American rivers to the Arctic Ocean is projected
to increase by 10 to 25% during the 21st century.
Recently, most of the statistical downscaling studies for
the Arctic have been performed for the European sector
as part of the Norwegian Regional Climate Development
Under Global Warming (RegClim) project or the Swedish
Regional Climate Modelling Programme (SWECLIM),
both of which use regional modeling and statistical downscaling to generate climate scenarios. Statistical downscaling using results from several global models and for various emissions scenarios has been completed (Benestad,
2002, 2004; Chen et al., 2001).The primary global model
used in the RegClim project was ECHAM4/OPYC3
(Roeckner et al., 1996, 1999), forced with the IS92a
emissions scenario.The primary case studied was GSDIO,
which included changes in GHGs, tropospheric ozone,
and direct as well as indirect sulfur aerosol forcing
(Roeckner et al., 1999). In SWECLIM, the main global
models were HadCM2 (Johns, 1996; Johns et al., 1997)
and ECHAM4/OPYC3. Hanssen-Bauer and Førland
(2001) evaluated the ECHAM4/OPYC3 simulation of
present-day climate over Norway and Svalbard, while
Räisänen and Döscher (1999) evaluated the HadCM2 simulation of the present-day climate of northern Europe.
Sea-level pressure has proven to be a good indicator for
the Scandinavian climate (Busuioc et al., 2001b; Chen,
2000; Chen and Hellström, 1999; Hanssen-Bauer and
Førland, 2001), and was therefore used as a large-scale
predictor in the statistical downscaling models.
Depending on predictands, additional predictors included air temperature at 2 m height, humidity, and precipitation.The models were developed by linear techniques:
CCA, SVD, or MLR. Monthly precipitation and temperatures at selected stations were the main predictands
(Benestad, 2002, 2004; Busuioc et al., 2001a; Chen and
Chen, 1999; Hanssen-Bauer et al., 2003; Hellström et
al., 2001). Although not used for scenario estimation,
some non-standard climate variables such as annual max-
Pre-release Version 24 March, 2005
www.acia.uaf.edu
138
Arctic Climate Impact Assessment
locations that are exposed to temperature inversions.
Hanssen-Bauer et al. (2003) argue that it is reasonable to
expect weaker winter inversions in the future, and that
greater winter warming rates should be expected in valleys compared to mountains.
Fig. 4.32. Cumulative frequency plots of observed and modeled
(a) mean annual temperature and (b) mean seasonal temperature
at Svalbard Airport for various time slices (Hanssen-Bauer, 2002).
imum sea ice extent over the Baltic Sea (Chen and Li,
2004; Omstedt and Chen, 2001), sea level near
Stockholm (Chen and Omstedt, 2002), and spring phytoplankton blooms in a Swedish lake (Blenckner and
Chen, 2003) were also linked to atmospheric circulation
and may thus be projected by statistical downscaling.
The statistically downscaled temperature scenario based
upon the GSDIO integration projects increases in mean
annual temperature of 0.2 to 0.5 ºC per decade in
Norway up to 2050 (Hanssen-Bauer et al., 2003), and
0.6 ºC per decade in Svalbard (Hanssen-Bauer, 2002).
Cumulative frequency (relative number of years that
temperatures go below a certain threshold given as a
function of the threshold value) plots of annual and seasonal mean temperatures based on this study (Fig. 4.32)
show good correspondence between observations and
model output for time slices within the 20th century.
The smallest warming rates were simulated in southern
Norway along the coast; the rates increase when moving
inland and northward. Along the coast of southern
Norway, the modeled warming rates are similar in all
seasons. Farther north and in the interior, considerably
larger warming rates are projected for winter than for
summer. Comparing these results to the results from
dynamic downscaling of the same integration (Bjørge et
al., 2000) shows only minor differences in summer and
autumn. In winter and spring, on the other hand, the
statistical downscaling projects greater warming rates for
Benestad (2002, 2004) shows that statistical downscaling
from several different climate models gives different local
warming rates over Fennoscandia. Still, the temperature
signal is robust in some respects: all models simulate
warming; the warming is larger inland than along the
coast; and the seasonal patterns are similar. A comparison
of statistically downscaled temperature scenarios for
Svalbard based on the GSDIO integration and HadCM3,
both forced with the IS92a emissions scenario, and
NCAR’s CSM forced with a 1% per year increase in
atmospheric CO2 concentration, revealed differences
(Fig. 4.33) that to a large degree can be explained in
terms of different descriptions of sea-ice extent
(Benestad et al., 2002).The HadCM3 model, which projects significantly stronger warming in this area than the
other simulations, projects a substantial retreat of sea ice
in the Barents Sea.The CSM model, which projects the
most moderate warming rates, shows no melting of sea
ice this area. Local temperature scenarios in the high
Arctic are thus closely related to the projected changes in
regional sea-ice cover. If the AOGCM fails to reproduce
either the present sea-ice border or future melting in the
region, the local temperature projections will be suspect.
Hellström et al. (2001) compared two dynamically and
statistically downscaled precipitation scenarios for
Sweden.The precipitation climates of the GCMs, dynamic
models (i.e., RCMs), and statistical models from the control runs were also compared with respect to their ability
to reproduce the observed seasonal cycle (Fig. 4.34).
Improvements in the representation of the seasonal cycle
by the downscaling models compared to the GCMs significantly increase the credibility of the downscaling models.
Fig. 4.33. Projections of seasonal and annual temperature
increases between 1981–2000 and 2030–2050 for the arctic stations Bjørnøya, Hopen, and Svalbard Airport, based on statistical
downscaling of three different general circulation models:
ECHAM4/OPYC3, HadCM3, and CSM (Benestad et al., 2002).
Pre-release Version 24 March, 2005
www.acia.uaf.edu
139
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
reduction in the annual frequency of days with precipitation of 1.4%. An increase in precipitation intensity is
projected throughout most of the year, especially during
winter.The major changes include a substantial increase
in winter precipitation, a delay in the timing of the summer maximum, and prolonged duration of the winter
maximum relative to present-day climate.This indicates
a more maritime precipitation climate in the scenario
climate compared to the control climate.
Fig. 4.34. Seasonal cycle of observed and modeled control
period (1921–1950) precipitation for Kvikkjokk in northern
Sweden (Hellström et al., 2001).
Chen et al. (2001) applied statistically downscaled scenarios from 17 CMIP2 AOGCMs to quantify AOGCMrelated uncertainty in the estimation of precipitation
scenarios.The result shows that there is an overall projected increase in annual precipitation over the 21st century throughout Sweden.The projected increase is
greater in northern than southern Sweden.The precipitation in autumn, winter, and spring is projected to
increase throughout the country, whereas decreasing
summer precipitation is projected for the southern part
of the country.The estimates for winter have a higher
level of confidence than the estimates for summer. A statistically downscaled precipitation scenario based upon
the GSDIO integration (Hanssen-Bauer et al., 2003) also
projects increased annual precipitation in Norway.The
projected rates of increase are smallest in southeastern
Norway, where they are not statistically significant, and
greatest along the northwestern and western coast,
where they are highly significant (Fig. 4.35). In winter
and autumn, statistically significant positive trends are
projected for most of Norway, while most of the modeled changes in spring and summer precipitation are not
statistically significant.
To date, statistical downscaling has primarily concentrated on monthly and annual scales. However, Linderson et
al. (2004) developed downscaling models for daily statistics based upon monthly precipitation values for southern Sweden. Future scenarios of selected daily precipitation statistics were downscaled from a GCM developed
at the Canadian Centre for Climate Modelling and
Analysis (CGCM1; e.g., Flato et al., 2000).The downscaling models use large-scale precipitation, relative
humidity, and circulation indices as predictors.The models are skillful in reproducing the variability of mean
precipitation and the frequency of days with no precipitation, but less skillful concerning extremes and statistics
of days with precipitation. By the time that atmospheric
CO2 doubles in the model, the CGCM1 projects an
increase of 10% in annual mean precipitation (statistically significant at the 95% level), and an insignificant
Statistically downscaled climate scenarios for the North
American and Russian parts of the Arctic have not yet
been published. However, the Canadian Climate Impacts
Scenarios (CCIS) Project has validated a downscaling
tool, the Statistical DownScaling Model (SDSM;Wilby
et al., 2002).The CCIS Project is providing predictor
variable data to support the SDSM.
4.6.3. Summary
Although only a few statistical downscaling studies have
been performed for arctic localities so far, results from the
available studies indicate that these methods are able to
resolve projected changes in temperature gradients
between the coast and inland, and changes in valley temperature inversions.They also generate a more realistic
representation of the annual precipitation cycle than the
AOGCMs. If careful attention is given to the choice of
predictors, scenarios from statistical downscaling and
regional modeling seem to be consistent. However, statis-
Precipitation change
(% per decade)
-1.5 – -1.25
-1.25 – -1.0
-1.0 – -0.75
-0.75 – -0.5
-0.5 – -0.25
-0.25 –
0
0
– 0.25
0.25 –
0.5
0.5 – 0.75
0.75 –
1.0
1.0 – 1.25
1.5
1.25 –
1.5 – 1.75
1.75 –
2.0
2.0 – 2.25
2.25 –
2.5
2.5 – 2.75
2.75 –
3.0
3.0 – 3.25
3.25 –
3.5
3.5 – 3.75
3.75 –
4.0
4.0 – 4.25
4.5
4.25 –
4.5 – 4.75
4.75 –
5.0
>5
No Data
Fig. 4.35. Projected change in annual precipitation in
Norway between 1961–1990 and 2031–2050 based on statistically downscaled output from the ECHAM4/OPYC3 GSDIO
scenario (Hanssen-Bauer et al., 2003).
Pre-release Version 24 March, 2005
www.acia.uaf.edu
140
Arctic Climate Impact Assessment
tical downscaling does not necessarily reproduce a physically based relationship between different climate elements. Conversely, the ability to downscale from several
models and integrations is useful for assessing uncertainty.
Results from downscaling temperature scenarios from
several models underscore that the local temperature projections in the high Arctic depend on how well changes in
sea-ice extent are represented in the AOGCMs.
4.7. Outlook for improving climate
change projections for the Arctic
To provide more reliable climate change scenarios for the
Arctic, several aspects of numerical climate models need
further development.The most challenging aspects of
model development are the physical parameterization
schemes: much of the uncertainty in Arctic climate
change projections can be attributed to an insufficient
knowledge of many of the physical processes active in the
Arctic domain.There is also substantial natural variability
in the arctic climate system and this part of the uncertainty cannot be eliminated simply by model development
and a refinement of the descriptions of physical processes.
The large-scale flow is dominated by variability patterns
such as the AO and the NAO (see section 2.2.2). In climate change simulations, both the frequency and nature
of these flow patterns may be altered.To assess changes in
the flow patterns, there must be a greater focus on the
climate predictability problem to probe the inevitable natural uncertainty through a systematic search in probability
space.To do this, ensemble projections are required
where both initial states and uncertain model parameters
are varied within a realistic range associated with a probability distribution.The development of more sophisticated
physical parameterization schemes and the introduction of
ensemble climate-change scenarios will both require considerable computing resources.
Historically, physical parameterization schemes have primarily been based on process descriptions and measurements from mid- and low latitudes (e.g., Randall et al.,
1998). Assuming that the same physical processes are relevant to the Arctic, the developments have “propagated”
from lower to higher latitudes in global models, and from
AOGCMs to RCMs. In recent years, the Arctic has
received particular attention from the climate modeling
community, motivated by the strong arctic response to an
increased GHG forcing in climate models.This has been
demonstrated in the northern high latitudes along with a
tremendous inter-model scatter, both in sensitivity to the
forcing and in simulating the observed climate in this
region. In particular, the amplification of global-model
systematic errors in regional arctic models presents a
serious challenge to future regional model developments.
In this section, research and model development priorities are summarized, aimed at an improvement of
AOGCM performance in the Arctic and, particularly, at
an increase in the credibility of AOGCM-based projections of future climate.
4.7.1.The Arctic part of the climate system –
a key focus in developing AOGCMs
Surface air temperature and precipitation are variables of
central interest from the viewpoint of AOGCM-based climate change scenarios.The level of confidence that can
be placed on the projected changes depends on the accuracy and adequacy of the representations of many physical processes, particularly boundary-layer fluxes of heat,
moisture, and momentum; clouds; and radiative fluxes.
Sea ice plays a dominant role in determining the intensity
of these fluxes in the Arctic and to a large extent determines the climate sensitivity of the Arctic, in particular to
GHG forcing. Description of sea ice is thus of central
importance in the arctic climate system, and there is a
considerable scope for improvement of the sea-ice components of current AOGCMs. More sophisticated treatments of sea-ice dynamics and thermodynamics can be
included – up to the level of stand-alone regional Arctic
Ocean/sea-ice models. However, even in the most comprehensive present-day sea-ice models, some important
processes are not properly represented, including heat
distribution between concurrent lateral and vertical melt
or growth of the ice and convective processes inherent in
sea ice (melt-pond and brine convection).
Improvements in the performance of AOGCM sea-ice
components are hampered by errors in the forcing fields
that determine sea-ice distribution. For example, the
systematic bias in the arctic surface atmospheric pressure
and the associated bias in the wind forcing of sea-ice as
simulated by atmospheric components of AOGCMs (and
stand-alone AGCMs) prevents even sea-ice models with
advanced dynamics from properly simulating spatial distributions of sea ice (Bitz et al., 2002).The causes of the
atmospheric pressure biases are not clear. Possible linkages include topographic (resolution) effects on atmospheric dynamics, lower boundary fluxes, as well as
atmospheric chemistry and dynamics of the upper atmosphere (Walsh, in press).
The atmospheric boundary layer in the Arctic is poorly
represented in current AOGCMs. It is unlikely that the
representation can be improved just by increasing model
vertical resolution. Insufficient understanding of the
physics of the atmospheric boundary layer in the Arctic
and the inappropriate parameterizations used in the current generation of AOGCMs call for further research in
this field.To a certain extent, the same can be said about
radiative transfer parameterizations, which should
account for specific features of the arctic atmosphere and
the underlying surface, including both the vertical and
the horizontal heterogeneity of this complex system
(Randall et al., 1998).
From a global perspective, clouds have been identified as
the most serious source of uncertainty in present-day
climate models (McAvaney et al., 2001).This is also true
for the Arctic. In particular, the multilayer arctic clouds
with their specific complexities associated with mixed
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
phases and low temperatures need to be represented
better. Other uncertain aspects of clouds involve the
radiative properties of ice crystals and the concentration
of various types of crystals, which have very different
properties with respect to their interaction with electromagnetic radiation.
Present climate-change modeling efforts largely focus
upon effects in the atmosphere, including effects on air
temperatures and precipitation. Modeling potential climate change in the marine Arctic has received less attention, although changes in the thermohaline circulation
have been extensively studied, primarily with low resolution, uncoupled models. Due to the lack of coordination among modeling studies, few definitive projections
can be made about changes to such variables as Arctic
Ocean temperatures and salinities, stratification, and circulation (including the thermohaline circulation). In
light of this, future modeling efforts should attempt to
more fully address changes in the ocean.This will
require better resolution in the ocean models and
improved coupling between the dynamic atmosphere
and dynamic ocean components, particularly in the presence of sea ice.
The freshwater budget of the Arctic Ocean (and its possible link to the intermittency of North Atlantic deepwater formation) is affected by the hydrological cycle
not only in the region, but also far beyond it, including
the vast terrestrial watersheds of the Arctic Ocean. For
satisfactory simulations, river discharge into the Arctic
Ocean needs to be properly represented in order to
maintain the observed stratification and sea-ice distribution and transport. It is not clear whether the simple
river discharge schemes used in current AOGCMs are
sufficient, although it appears that incorporating more
comprehensive river routing schemes will help ensure
proper seasonality of the discharge and result in an
improvement in the representation of Arctic Ocean general circulation. Accounting for the freshwater influx
into the ocean from glaciers and the Greenland Ice Sheet
will require more advanced parameterizations than those
employed today and, ideally, require introducing dynamics into the ice-sheet components.
Processes and feedbacks associated with vegetation may
also play an important role in the terrestrial Arctic,
affecting heat, water, and momentum fluxes.The effects
of vegetation on terrestrial snow cover and surface albedo, evapotranspiration processes, and the possible expansion of boreal forests into regions currently occupied by
tundra are among many processes that may potentially
be crucial in the context of climate change. Developing
comprehensive interactive dynamic vegetation components of AOGCMs should eventually increase confidence
in AOGCM-based projections of future climate.
Climatic changes of special concern for indigenous communities include weather variability and predictability;
the extent, thickness, and quality of sea ice; the extent,
duration, and hardness of snow cover; freeze-thaw cycles
141
(particularly in autumn, when a layer of ice on the
ground may be produced and last all winter, blocking
access to forage for grazing animals); sudden changes in
wind direction; and changes in the strength and frequency of winds and storms (Chapter 3).While most of these
quantities are either directly available, or easily derivable
from standard model outputs, representation of a few of
these variables will require additional efforts from the
modeling community in the future.
4.7.2. Improved resolution of arctic processes
To model climatically important processes in the Arctic,
models with a high spatial resolution are required.To
achieve this with present-day computing resources,
regional models are required. In the future, global models
may also have adequate resolution, but for the foreseeable
future regional models will be required to complement
the global simulations, because their results are closer to
actual local climatic conditions and can more easily be
translated into impacts than global model results.
When nested within a GCM, the large-scale circulation is
imposed by the lateral boundaries of the RCM. Regional
models are not able to, nor intended to, correct largescale errors made by the global model from which conditions are drawn.The role of the regional model is to add
regional detail and fine spatial and temporal scales to the
simulation, not to improve the large-scale simulation. An
alternative to regional models is presented by the evolving global variable-resolution stretched-grid approach
that provides additional spatial detail over a region of
interest (e.g., Giorgi et al., 2001).This technique allows
for a feedback to the global scale from the region with
high resolution.While this may seem appealing at first, it
raises the question of whether this feedback is preferable
when similar feedbacks from other regions are represented at a lower resolution.
The need for high resolution is not restricted to the
atmospheric model component.To simulate the coupled
atmosphere–ice–ocean system in the Arctic, a highresolution ocean component is also required. In particular, the coupling processes occur on small horizontal and
vertical scales, thus a high-resolution regional coupled
atmosphere-ocean model is needed. Some early versions
of such coupled models already exist, but much additional development work is required.
A further increase in atmospheric resolution (to <10 km
horizontally) will require the use of non-hydrostatic
model equations. New parameterizations of physical
processes such as cloud formation and turbulence are also
necessary at these scales.With a very high resolution
(<1 km), non-hydrostatic models start to resolve individual clouds, thus necessitating further changes in cloud
parameterizations. A special emphasis needs to be put on
cloud microphysics, including the ice crystals and
aerosols that provide nuclei for the condensation process.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
142
Arctic Climate Impact Assessment
The arctic climate depends on the unique high-latitude
characteristics of processes such as ice dynamics and
persistent low-level clouds. Simulation deficiencies are
partly due to coarse model resolution and partly due to
inadequate model process descriptions. As mentioned
previously, most model formulations are based on lowlatitude observations that do not cover the extreme conditions occurring in the Arctic.To validate coupled highresolution models in the Arctic, improved and extended
observational datasets are required. In situ observations
exist for a few locations and restricted time periods, but
more such datasets are needed.To obtain better coverage
in space and time, remote sensing instruments are necessary. Several satellite missions are planned that hopefully will provide observational datasets with a much
better coverage.
4.7.3. Better representation of the
stratosphere in AGCMs
Most current AGCMs are aimed at simulating tropospheric processes, and the stratosphere is only included
with a limited resolution. On the other hand, many of
the middle atmosphere three-dimensional circulation
models describe only the stratosphere and the mesosphere, having a lower boundary at the tropopause
(IPCC, 2001). Such models are primarily intended to
simulate processes that are internal to the stratosphere,
and it is assumed that the interaction with the troposphere can be neglected.
To model current arctic climate and stratospheric and
tropospheric ozone concentrations, as well as to project
their future changes, AGCMs must describe the troposphere and the stratosphere in comparable detail. Most
models assume that all ozone-related processes are located in the stratosphere: ozone and ozone-related species,
as well as their photochemical sources and sinks and air
transport in the ozone layer (20–30 km average height).
However, some of the stratospheric transport features
have a tropospheric origin.Two important processes in
this regard are planetary wave propagation in the northern mid-latitudes and gravitational wave destruction in
the middle and upper stratosphere.While the planetary
waves are well-resolved by climate models, gravity-wave
drag occurs at small scales and is therefore difficult to
simulate with the coarse grids of most current AGCMs.
In many AGCMs, these dynamic factors are roughly
parameterized by Rayleigh friction at upper model levels.This parameterization also serves the purpose of preventing spurious reflection of vertically propagating
gravity waves at the upper boundary.This feature is necessary in a climate model, but in order to resolve vertically propagating waves realistically, more resolution is
needed in the middle and upper stratosphere. Austin et
al. (1997) demonstrated that the shift from 19 to 49 levels in an AGCM with coupled chemistry considerably
improved the ozone and temperature simulation in the
winter stratosphere.
The AO is another important feature affecting the stratosphere (Wallace, 2000).The AO is a naturally occurring
phenomenon but difficult to project with current
GCMs. A gradual increase in the AO positive phase persistence has been observed in recent years (Hoerling et
al., 2001; Shindell et al., 2001).While the change may
be a natural fluctuation in the AO, it may also be a result
of increased atmospheric GHG concentrations. A better
resolution of the stratosphere in models is required to
determine whether this is the case, and, if it is, whether
further increases in GHG concentrations are likely to
exert a greater influence on the AO. A change in the AO
would also influence the ozone distribution in the arctic
stratosphere, giving rise to additional climate-relevant
feedbacks in the Arctic. One example could be a change
in the latitudinal heating gradient in the stratosphere
caused by a change in the ozone distribution. An altered
heating gradient would result in a changed temperature
gradient, which in turn would change the zonal wind
distribution.The zonal wind distribution determines the
vertical planetary wave propagation characteristics that
in turn affect ozone distribution.
4.7.4. Coupling chemical components to GCMs
Ozone is an important GHG and moderates fluxes of
ultraviolet radiation at ground level. In addition to an
adequate description of dynamic processes, GCMs must
incorporate detailed photochemical components for better simulation of ozone formation and destruction in the
atmosphere. Due to the complicated character of ozone
photochemistry in the arctic stratosphere, which has significant input from heterogeneous reactions on polar
stratospheric cloud (PSC) particle surfaces, the inclusion
of the microphysics of particle formation and destruction must be considered.This is omitted in the photochemical components of most present-day GCMs.
Instead, the observed spectra of PSC particles is assumed
to appear immediately when the air temperature drops
below a certain threshold temperature and to disappear
at once when the temperature rises above the threshold.
The actual delay in the observed PSC effects compared
to those modeled indicates the importance of considering PSC microphysics in models, especially in the simulation of arctic ozone “mini-holes” and their rapid evolution in space and time (Austin et al., 2003).
The denitrification of cold polar air in winter is another
process in the microphysics of PSC formation and the
chemistry that activates ozone-depleting chlorine radicals and repartitioning of bromine species. Polar stratospheric cloud particles that contain liquid nitric acid are
supercooled ternary solutions of nitric acid, sulfuric
acid, and water.They grow in the stratosphere to nitric
acid dihydrate (NAD) and nitric acid trihydrate (NAT)
large particles, which remove nitric acid from the stratosphere by gravitational sedimentation and contribute to
the denoxification (removal of nitric acid) of the arctic
stratosphere. Both NAD and NAT particles are formed
intensively at temperatures of 190 to 192 K – the
“nucleation window” (Tabazadeh et al., 2001).These
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
“window temperature” belts are persistent at the periphery of winter polar vortices in the Antarctic for several
months, and in the Arctic for about a month, and produce significant denitrified stratospheric layers
(Tabazadeh et al., 2001).
143
tainty aspects to be addressed thus include natural variability, uncertainties in model sensitivity to prescribed
forcings, and uncertainties in the forcings.
Another atmospheric chemistry aspect of the Arctic is
the production of cloud condensation nuclei near the
surface and the possible involvement of naturally occurring dimethyl sulfide (DMS) in this process. Dimethyl
sulfide particles originate from arctic seawater, and the
flux to the atmosphere is thus strongly coupled to the
existence of sea ice. It may be that the local arctic production of DMS is a determining factor for droplet size
distributions in low clouds and thus may have a significant effect on low-cloud radiative properties. If this is
the case, cloud properties would be sensitive to the
occurrence of sea ice and a dramatic change in sea-ice
distribution would affect the arctic radiative balance.
This type of effect, as well as arctic haze effects, whose
radiative forcing has been estimated from observations
(e.g., Herber et al., 2002; Quinn et al., 2002), need to
be included in AGCMs.
Estimates of extreme events and their frequency of
occurrence also require ensemble simulations. For precipitation in particular, extreme events are often more
interesting than changes in the mean (Palmer and
Räisänen, 2002).To obtain reliable estimates of changes
in the frequency of extreme events, ensemble simulations are necessary.This is thus an added benefit of climate-change projection ensembles and is required to
make projections of changes in storm frequencies or
other extreme events. It has recently been shown that an
increased GHG forcing could contribute to an increase
in intense storm events in particular areas of the North
Atlantic Ocean and Western Europe (Van den Brink et
al., 2003). At the same time, a decrease in storm frequencies is projected for other regions of the North
Atlantic.To arrive at this result, a very large ensemble
was used, and in order to achieve that, a simplified
atmospheric model (quasi-geostrophic, three vertical
levels, and a coarse horizontal resolution) had to be utilized.The drawback of using such a simplified model is
that storm dynamics are not described in full detail and
storm characteristics have to be derived from empirically based, statistical methods similar to the downscaling
technique discussed in section 4.6. Other studies using
more advanced model tools but smaller ensembles have
not been able to simulate significant increases in storm
frequencies (Carnell and Senior, 1998; Knippetz et al.,
2000; Lunkeit et al., 1996; Ulbrich and Christoph,
1999).
4.7.5. Ensemble simulations
4.7.6. Conclusions
A more ambitious strategy for ensemble climate simulations is needed in order to better understand natural climate variability in the Arctic and how it may be affected
by global climate change. In discussing the impacts of
climate change, changes in the distribution of climatic
events are as interesting as changes in the mean.The
ACIA used the results of five climate models to study
future changes in arctic climate. In order to increase the
accuracy of the different error estimates, a larger scenario sample is needed. In numerical weather prediction, experience has shown that a sample involving 50 to
100 simulations with identical models but different initial states gives a reasonable estimate of forecast uncertainties. For arctic climate change, error estimates based
on a sample of this size could be adequate. In addition, it
would be advantageous to increase model resolution to
better capture physical processes and to better describe
sharp spatial gradients (fronts), which are often the
regions where extreme events occur. Both types of
improvement require large additional computing
resources. Further research is needed to find a reasonable balance between ensemble size, model resolution,
and the complexity of physical process descriptions. For
climate simulation ensembles, it is also necessary to perturb model parameters and external forcings.The uncer-
The general increase in computing resources that have
become available for climate system modeling in recent
years favors progress in developing new generations of
AOGCMs – mostly by adding new components, increasing resolution, and extending ensembles of simulations.
Conversely, the Arctic is one of the regions of the world
with limited availability of observational data necessary
for model validation and evaluation (e.g.,Walsh, in
press). Nevertheless, it has been shown that model performance can be improved with systematic model
improvements and better resolution. For the Arctic, it is
necessary to perform climate change simulations involving the entire globe; however, spatial resolution in the
Arctic could be improved with the use of regional models driven by global simulations.The ultimate goal is to
use as high a resolution as possible over the entire globe.
In simulating arctic climate change, sea-ice processes are
of primary importance. Boundary-layer fluxes and
clouds are closely linked with sea-ice processes. All components require improvements to increase confidence in
climate change projections. Expectations for model
improvement are increasing because of increasing international activity in the field of model intercomparison
exercises (e.g., Puri, 2002; Box 4.2), allowing the iden-
This phenomenon as well as the PSC microphysics and
chemistry have spatial and temporal scales finer than
current GCM and CTM grids can resolve. A suitable
parameterization of these effects is needed in addition to
an elaboration of the whole photochemical computation
scheme.Together with the necessary refinement of the
simulation of dynamic processes, these requirements
make the problem of arctic ozone modeling computationally demanding and scientifically challenging.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
144
Arctic Climate Impact Assessment
tification of model errors, their causes, and how they
may be reduced.
Some scientists doubt that AOGCMs can provide realistic scenarios of future climate change. However, even if
present day models have major shortcomings and need
to be improved, they still provide useful information
about possible changes in the future climate.The models
are based on a physical understanding of the climate system and, as such, provide a physically coherent picture
of likely climate change.There are very few other methods, if any, which can be used to provide such credible
climate change estimates. Statistical methods, other than
simple extrapolation of present trends, require a physical
model in the background to provide a basis to generate
statistically representative estimates of variables that cannot be deduced directly from the physical model.The
authors of this chapter are thus confident that future
model improvements will provide better estimates of the
arctic climate change that may occur as a result of
increasing atmospheric GHG concentrations.
There will always be uncertainties in the estimates and
some of these uncertainties cannot be reduced below a
certain level.These include, for example, uncertainties
associated with the lack of observations to provide an
accurate initial state for a model simulation, model
parameter uncertainties, and the inherent limited predictability of any atmospheric/oceanic simulation
(Lorenz, 1963).While the level of uncertainty can be
lowered, it will never be certain that all physical
processes relevant to climate change have been included
in a model simulation.There could still be surprises to
come in the understanding of climate change. Solar variability, the effects of cosmic rays, and volcanic eruptions
may all contribute more to arctic climate change than is
presently thought, but this remains to be seen. As climate change science progresses there will always be new
results that could significantly change understanding of
how the arctic climate system works; however, the present estimates are based on the best knowledge available
today about climate change.
References
Austin, J., N. Butchart and R. Swinbank, 1997. Sensitivity of ozone and
temperature to vertical resolution in a GCM with coupled stratospheric chemistry. Quarterly Journal of the Royal Meteorological
Society, 123:1405-1431.
Austin, J., D. Shindell, S.R. Beagley, C. Bruhl, M. Damers, E. Manzini,
T. Nagashima, P. Newman, S. Pawson, G. Pitari, E. Rozanov, C.
Schnadt and T.A. Shepherd, 2003. Uncertainties and assessments of
chemistry-climate models of the stratosphere. Atmospheric
Chemistry and Physics, 3:1-27.
Bailey, D.A. and A.H. Lynch, 2000a. Development of an Antarctic
Regional Climate System Model: Part 1. Sea ice and large-scale circulation. Journal of Climate, 13:1337-1350.
Bailey, D.A. and A.H. Lynch, 2000b. Development of an Antarctic
Regional Climate System Model: Part 2. Station validation and surface energy balance. Journal of Climate, 13:1351-1361.
Bailey, D.A., A.H. Lynch and K.S. Hedström, 1997.The impact of ocean
circulation on regional polar climate simulations using the Arctic
Region Climate System Model. Annals of Glaciology, 25:203-207.
Barthelet, P., L.Terray and S.Valcke, 1998.Transient CO2 experiment
using the ARPEGE/OPAICE non flux corrected coupled model.
Geophysical Research Letters, 25:2277-2280.
Benestad, R., 2002. Empirically downscaled multi-model ensemble temperature and precipitation scenarios for Norway. Journal of Climate,
15:3008-3027.
Benestad, R., 2004.Tentative probabilistic temperature scenarios for
northern Europe.Tellus A: Dynamic Meteorology and
Oceanography, 56:89-101.
Benestad, R., E.J. Forland and I. Hanssen-Bauer, 2002. Empirically
downscaled temperature scenarios for Svalbard. Atmospheric Science
Letters, 3(2-4):71-93.
Bengtsson, L.,V.A. Semenov and O. Johannessen, 2003.The Early
Century Warming in the Arctic – A Possible Mechanism. Rep. 345.
Max Planck Institute for Meteorology, Hamburg, Germany, 31pp.
Bitz, C., G. Flato and J. Fyfe, 2002. Sea ice response to wind forcing
from AMIP models. Journal of Climate, 15:523-535.
Bjørge, D., J.E. Haugen and T.E. Nordeng, 2000. Future Climate in
Norway. Dynamical downscaling experiments within the RegClim project. Research Report 103, Norwegian Meteorological Institute, Oslo.
Blenckner,T. and D. Chen, 2003. Comparison of the impact of regional
and north-Atlantic atmospheric circulation on an aquatic ecosystem.
Climatic Research, 23:131-136.
Boer, G.J., 2000a. Analysis and verification of model climate. In: P. Mote
and A. O’Neill (eds.). Numerical Modeling of the Global
Atmosphere in the Climate System. NATO Science Series C-550.
Kluwer Academic Publishers.
Boer, G.J., 2000b. Climate model intercomparison. In: P. Mote and A.
O’Neill (eds.). Numerical Modeling of the Global Atmosphere in the
Climate System. NATO Science Series C-550. Kluwer Academic
Publishers.
Boer, G.J., N.A. McFarlane and M. Lazare, 1992. Greenhouse gasinduced climate change simulated with the CCC second-generation
general circulation model. Journal of Climate, 5:1045-1077.
Boville, B.A. and P.R. Gent, 1998.The NCAR Climate System Model,
Version One. Journal of Climate, 11:1115-1130.
Boville, B.A., J.T. Kiehl, P.J. Rasch and F.O. Bryan, 2001. Improvements
to the NCAR CSM-1 for transient climate simulations. Journal of
Climate, 14:164-179.
Braconnot, P. (ed.), 2000. Paleoclimate Modeling Intercomparison
Project (PMIP): Proceedings of the Third PMIP Workshop, La
Huardière, Canada, 4-8 October 1999. ICPO Publications Series
No.34, PAGES 2000-1,WMO/TD No. 1007,WCRP Series Report
No. 111, 271pp.
Braconnot, P., O. Marti and S. Joussaume, 1997. Adjustment and feedbacks in a global coupled ocean-atmosphere model. Climate
Dynamics, 13:507-519.
Broecker,W.S., 1997.Thermohaline circulation, the achilles heel of our
climate system: will man-made CO2 upset the current balance?
Science, 278:1582-1588.
Bromwich, D., J. Cassano,T. Klein, G. Heinemann, K. Hines, K. Steffen
and J. Box, 2001. Mesoscale modeling of katabatic winds over
Greenland with the Polar MM5. Monthly Weather Review,
129(9):2290-2309.
Bryan, K., S. Manabe and R.C. Pacanowski, 1975. A global ocean-atmosphere climate model. II: the oceanic circulation. Journal of Physical
Oceanography, 5:30-46.
Bryazgin, N.N., 1976.Yearly mean precipitation in the Arctic region
accounting for measurement errors. Proceedings of the Arctic and
Antarctic Research Institute, 323:40-74. (In Russian)
Busuioc, A., H. von Storch and R. Schnur, 1999.Verification of GCMgenerated regional seasonal precipitation for current climate and of
statistical downscaling estimates under changing climate conditions.
Journal of Climate, 12:258-272.
Busuioc, A., D. Chen and C. Hellström, 2001a. Performance of statistical downscaling models in GCM validation and regional climate estimates: application for Swedish precipitation. International Journal of
Climatology, 21:557-578.
Busuioc, A., D. Chen and C. Hellström, 2001b.Temporal and spatial
variability of precipitation in Sweden and its link with the large scale
atmospheric circulation.Tellus, 53A(3):348-367.
Carnell, R.E. and C.A. Senior, 1998. Changes in mid-latitude variability
due to increasing greenhouse gases and sulphate aerosols. Climate
Dynamics, 14:369-383.
Carter,T.R., M.L. Parry, H. Harasawa and S. Nishioka, 1994. IPCC
Technical Guidelines for Assessing Climate Change Impacts and
Adaptations. Department of Geography, University College, London.
Carter,T.R., M. Hulme, J.F. Crossley, S. Malyshev, M.G. New, M.E.
Schlesinger and H.Tuomenvirta, 2000. Climate Change in the 21st
Century – Interim Characterizations based on the New IPCC
Emissions Scenarios.The Finnish Environment 433, Finnish
Environment Institute, Helsinki, 148pp.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Carter,T.R., E.L. La Rovere, R.N. Jones, R. Leemans, L.O. Mearns, N.
Naki5enovi5, A.B. Pittock, S.M. Semenov and J. Skea, 2001.
Developing and applying scenarios. In: J.J. McCarthy, O.F. Canziani,
N.A. Leary, D.J. Dokken and K.S.White (eds.). Climate Change
2001: Impacts, Adaptation, and Vulnerability. Pp. 145-190.
Contribution of Working Group II to the Third Assessment Report of
the Intergovernmental Panel on Climate Change. Cambridge
University Press.
Cassano, J.J.,T.R. Parish and J. King, 2001. Evaluation of turbulent surface flux parameterizations for the stable boundary layer over Halley,
Antarctica. Monthly Weather Review, 129:26-46.
Cavazos,T., 1999: Large-scale circulation anomalies conductive to extreme
precipitation events and derivation of daily rainfall in northeastern
Mexico and southeastern Texas. Journal of Climate, 12:1506-1523.
Chen, D., 2000. A monthly circulation climatology for Sweden and its
application to a winter temperature case study. International Journal
of Climatology, 20:1067-1076.
Chen, D. and Y. Chen, 1999. Development and verification of a multiple
regression downscaling model for monthly temperature in Sweden. In:
D. Chen, C. Hellström,Y. Chen (eds.). Preliminary Analysis and
Statistical Downscaling of Monthly Temperature in Sweden, pp. 41-55.
Report C16, Earth Sciences Centre, University of Gothenburg, Sweden.
Chen, D., and Y. Chen, 2003. Association between winter temperature
in China and upper air circulation over East Asia revealed by
Canonical Correlation Analysis. Global and Planetary Change,
37:315-325.
Chen, D. and C. Hellström, 1999.The influence of the North Atlantic
Oscillation on the regional temperature variability in Sweden: spatial
and temporal variations.Tellus, 51A(4):505-516.
Chen, D. and X. Li, 2004. Scale dependent relationship between maximum ice extent in the Baltic Sea and atmospheric circulation. Global
and Planetary Change, 41:275-283.
Chen, D. and A. Omstedt, 2002. Using statistical downscaling to quantify
the GCM-related uncertainty in regional climate change scenarios: A
case study of Swedish precipitation. SWECLIM Newsletter, Swedish
Meteorological and Hydrological Institute, Norrköping, Sweden.
Chen, D., C. Achberger, J. Räisänen and C. Hellström, 2001. Using statistical downscaling to quantify the GCM-related uncertainty in
regional climate change scenarios: A case study of Swedish precipitation. SWECLIM Newsletter, Swedish Meteorological and
Hydrological Institute, Norrköping, Sweden.
Christensen, J.H. and O.B. Christensen, 2003. Severe summer flooding
in Europe. Nature, 421:805-806.
Christensen, J.H. and P. Kuhry, 2000. High resolution regional climate
model validation and permafrost simulation for the East-European
Russian Arctic. Journal of Geophysical Research, 105:29647-29658.
Christensen, J.H., O.B. Christensen, P. Lopez, E. van Meijgaard and M.
Botzet, 1996.The HIRHAM4 Regional Atmospheric Climate Model.
Danish Meteorological Institute, Scientific Report 96-4.
Christensen, J.H., B. Machenhauer, R.G. Jones, C. Schär, P.M. Ruti, M.
Castro and G.Visconti, 1997.Validation of present-day regional climate simulations over Europe: LAM simulations with observed
boundary conditions. Climate Dynamics, 13:489-506.
Christensen, J.H.,T.R. Carter and F. Giorgi, 2002. PRUDENCE
employs new methods to assess European climate change. Eos,
Transactions, American Geophysical Union, 83(13):147.
Christensen, O.B., 1999. Relaxation of soil variables in a Regional
Climate Model.Tellus, 51A:674-685.
Clair,T.A. and J. Ehrman, 1998. Using neural networks to assess the
influence of changing seasonal climates in modifying discharge, dissolved organic carbon, and nitrogen export in eastern Canadian
rivers.Water Resources Research, 34(3):447-455.
Clair,T.A., J. Ehrman and K. Higuchi, 1998. Changes to the runoff of
Canadian ecozones under a doubled CO2 atmosphere. Canadian
Journal of Fisheries and Aquatic Sciences, 55(11):2464-2477.
Claussen, M., L.A. Mysak, A.J.Weaver, M. Crucifix,T. Fichefet, M.-F.
Loutre, S.L.Weber, J. Alcamo,V.A. Alexeev, A. Berger, R. Calov, A.
Ganopolski, H. Goosse, G. Lohmann, F. Lunkeit, I.I. Mokhov,V.
Petoukhov, P. Stone and Z.Wang, 2002. Earth system models of
intermediate complexity: closing the gap in the spectrum of climate
system models. Climate Dynamics, 18:579-586.
Conway, D. and P.D. Jones, 1998.The use of weather types and air flow
indices for GCM downscaling. Journal of Hydrology, 212:348-361.
Corte-Real, J., X. Zhang and X.Wang, 1995. Large-scale circulation
regimes and surface climate anomalies over the Mediterranean.
International Journal of Climatology, 15:1135-1150.
Crane, R.G. and B.C. Hewitson, 1998. Doubled CO2 precipitation changes
for the Susquehanna Basin: downscaling from the genesis General
Circulation Model. International Journal of Climatology, 18:65-76.
145
Cubasch, U., K. Hasselmann, H. Höck, E. Maier-Raimer, U.
Mikolajewicz, B.D. Santer and R. Sausen, 1992.Time-dependent
greenhouse warming computations with a coupled ocean-atmosphere
model. Climate Dynamics, 9:55-69.
Cubasch, U., H. von Storch, J.Waszkewitz and E. Zorita, 1996.
Estimates of climate change on southern Europe derived from
dynamical climate model output. Climate Research, 7:129-149.
Cubasch, U., G.A. Meehl, G.J. Boer, R.J. Stouffer, M. Dix, A. Noda, C.A.
Senior, S. Raper and K.S.Yap, 2001. Projections of future climate
change. In: J.T. Houghton,Y. Ding, D.J. Griggs, M. Noguer, P.J. van der
Linden, X. Dai, K. Maskell and C.A. Johnson (eds.). Climate Change
2001:The Scientific Basis, pp. 526-582. Contribution of Working Group
I to the Third Assessment Report of the Intergovernmental Panel on
Climate Change. Cambridge University Press.
Curry, J.A. and A.H. Lynch, 2002. Comparing Arctic Regional Climate
Models. Eos,Transactions, American Geophysical Union, 83:87.
Curry R., B. Dickson and I.Yashayaev, 2003. A change in the freshwater
balance of the Atlantic Ocean over the past four decades. Nature,
426:826-829.
Delworth,T.L. and T.R. Knutson, 2000. Simulation of early 20th century global warming. Science, 287:2246-2250.
Delworth,T.L., R.J. Stouffer, K.W. Dixon, M.J. Spelman,T.R. Knutson,
A.J. Broccoli, P.J. Kushner and R.T.Wetherald, 2002. Simulation of
climate variability and change by the GFDL R30 coupled climate
model. Climate Dynamics, 19(7):555-574.
Denis, B., R. Laprise and D. Cava, 2002. Downscaling ability of one-way
nested regional climate models: the big-brother experiment. Climate
Dynamics, 18:627-646.
Dethloff, K., A. Rinke, R. Lehmann, J.H. Christensen, M. Botzet and B.
Machenhauer, 1996. A regional climate model of the Arctic atmosphere. Journal of Geophysical Research, 101:23401-23422.
Dethloff, K., C. Abegg, A. Rinke, I. Hebestadt and V.F. Romanov, 2001.
Sensitivity of Arctic climate simulations to different boundary-layer
parameterizations in a regional climate model.Tellus, 53A:1-26.
Dethloff, K., M. Schwager, J.H. Christensen, S. Kiilsholm, A. Rinke,W.
Dorn, F. Jung-Rothenhäusler, H. Fischer, S. Kipfstuhl and H. Miller,
2002. Greenland precipitation from ice core estimates and regional
climate model simulations. Journal of Climate, 15:2821-2832.
Dickson, B., I.Yashayaev, J. Meincke, B.Turrell, S. Dye and J. Holfort,
2002. Rapid freshening of the deep North Atlantic Ocean over the
past four decades. Nature, 416:832-837.
Dorn,W., K. Dethloff, A. Rinke and E. Roeckner, 2003. Competition of
NAO regime changes and increasing greenhouse gases and aerosols
with respect to Arctic climate projections. Climate Dynamics, 21(56): 447-458.
Ehrman, J., K. Higuchi and T.A. Clair, 2000. Backcasting to test the use
of neural networks for predicting runoff in Canadian rivers. Canadian
Water Resources Journal, 25(3):279-291.
Emori, S.,T. Nozawa, A. Abe-Ouchi, A. Numaguti, M. Kimoto and T.
Nakajima, 1999. Coupled ocean-atmosphere model experiments of
future climate change with an explicit representation of sulfate
aerosol scattering. Journal of the Meteorological Society of Japan,
77:1299-1307.
Enke,W. and A. Spekat, 1997. Downscaling climate model outputs into
local and regional weather elements by classification and regression.
Climate Research, 8:195-207.
Etchevers, P., E. Martin, R. Brown, C. Fierz,Y. Lejeune, E. Bazile, A.
Boon,Y.-J. Dai, R. Essery, A. Fernandez,Y. Gusev, R. Jordan,V.
Koren, E. Kowalczyck, R. Nasonova, D. Pyles, A. Schlosser, A.
Shmakin,T.G. Smirnova, U. Strasser, D.Verseghy,T.Yamazaki and Z.L.Yang, 2002. SnowMIP, an intercomparison of snow models: first
results. In: Proceedings of the International Snow Science Workshop,
Penticton, Canada, 29 September - 4 October 2002, 8pp.
Flato, G.M. and G.J. Boer, 2001.Warming asymmetry in climate change
experiments. Geophysical Research Letters, 28:195-198.
Flato, G.M., G.J. Boer,W.G. Lee, N.A. McFarlane, D. Ramsden, M.C.
Reader and A.J.Weaver, 2000.The Canadian Centre for Climate
Modelling and Analysis global coupled model and its climate. Climate
Dynamics, 16:451-467.
Folland, C., J. Shukla, J. Kinter and M. Rodwell, 2002.The climate of
the twentieth century project. Exchanges (Newsletter of CLIVAR),
7(2):37-39.
Frei, A., J.A. Miller and D.A. Robinson, 2003a. Improved simulations of
snow extent in the second phase of the Atmospheric Model
Intercomparison Project (AMIP-2). Journal of Geophysical Research,
108:(D12), 4369, doi:10.1029/2002JD003030.
Frei, C., J.H. Christensen, M. Deque, D. Jacob, R.G. Jones and P.L.
Vidale, 2003b. Daily precipitation statistics in regional climate models: evaluation and intercomparison for the European Alps. Journal of
Geophysical Research, 108: 10.1029/2002JD002287.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
146
Arctic Climate Impact Assessment
Furevik,T., M. Bentsen, H. Drange, I.K.T. Kindem, N.G. Kvamstø and
A. Sorteberg, 2003. Description and validation of the Bergen
Climate Model: ARPEGE coupled with MICOM. Climate Dynamics,
21:27-51.
Ganopolski, A. and S. Rahmstorf, 2001. Rapid changes of glacial climate
simulated in a coupled climate model. Nature, 409:153-158.
Gates,W.L., 1992.AMIP:The Atmospheric Model Intercomparison Project.
Bulletin of the American Meteorological Society, 73:1962-1970.
Gates,W.L., J.S. Boyle, C.C. Covey, C.G. Dease, C.M. Doutriaux, R.S.
Drach, M. Fiorino, P.J. Gleckler, J.J. Hnilo, S.M. Marlais,T.J.
Phillips, G.L. Potter, B.D. Santer, K.R. Sperber, K.E.Taylor and D.N.
Williams, 1998. An overview of the results of the Atmospheric
Model Intercomparison Project (AMIP). PCMDI Rep. No. 45,
UCRL-JC-129928. Program for Climate Model Diagnosis and
Intercomparison, Lawrence Livermore National Laboratory,
Livermore, CA, 29 pp. + figs.
Gibson, J.K., P. Kållberg, S. Uppala, A. Hernandez, A. Nomura and E.
Serrano, 1997. ERA description. ECMWF Reanalysis Project Rep.
Series 1. European Centre for Medium Range Weather Forecasts,
Reading, UK, 66pp.
Giorgi, F. and L.O. Mearns, 1999. Introduction to special section:
Regional climate modeling revisited. Journal of Geophysical
Research, 104:6335-6352.
Giorgi, F. and L.O. Mearns, 2003. Probability of regional climate
change calculated using the Reliabibility Ensemble Averaging (REA)
method. Geophysical Research Letters, 30(12):311-314.
Giorgi, F., B. Hewitson, J. Christensen, M. Hulme, H. von Storch, P.
Whetton, R. Jones, L. Mearns and C. Fu, 2001. Regional climate
information – evaluation and projections. In: J.T. Houghton,Y. Ding,
D.J. Griggs, M. Noguer, P.J. van der Linden, X. Dai, K. Maskell and
C.A. Johnson (eds.). pp. 583-638. Climate Change 2001:The
Scientific Basis. Contribution of Working Group I to the Third
Assessment Report of the Intergovernmental Panel on Climate
Change. Cambridge University Press.
Goodess, C.M. and J.P. Palutikof, 1998. Development of daily rainfall
scenarios for southeast Spain using a circulation-type approach to
downscaling. International Journal of Climatology, 10:1051-1083.
Gordon, C., C. Cooper, C.A. Senior, H. Banks, J.M. Gregory,T.C.
Johns, J.F.B. Mitchell and R.A.Wood, 2000.The simulation of SST,
sea ice extents and ocean heat transports in a version of the Hadley
Centre coupled model without flux adjustments. Climate Dynamics,
16:147-168.
Hagemann, S. and L. Dümenil, 1998. A parameterization of the lateral
water flow for the global scale. Climate Dynamics, 14:17-31.
Hahn, C.J., S.G.Warren and J. London, 1995.The effect of moonlight
on observations of cloud cover at night, and applications to cloud climatology. Journal of Climate, 8:1429-1466.
Hanna, E. and P.Valdes, 2001.Validation of ECMWF (re)analysis surface
climate data, 1979-1998 for Greenland and implications for mass balance modelling of the ice sheet. International Journal of Climatology,
21(2):171-195.
Hanssen-Bauer, I., 2002.Temperature and precipitation at Svalbard
1900-2050: measurements and scenarios. Polar Record,
38(206):225-232.
Hanssen-Bauer, I. and E.J. Førland, 2001.Verification and analysis of a
climate simulation of temperature and pressure fields over Norway
and Svalbard. Climate Research, 16:225-235.
Hanssen-Bauer, I., E.J. Førland, J.E. Haugen and O.E.Tveito, 2003.
Temperature and precipitation scenarios for Norway: Comparison of
results from dynamical and empirical downscaling. Climate Research,
25:15-27.
Harding, R.J., S.-E. Gryning, S. Halldin and C.R. Lloyd, 2001. Progress
in understanding of land surface/atmosphere exchanges at high latitudes.Theoretical and Applied Climatology, 70:5-18.
Haugen, J.E., D. Bjørge and T.E. Nordeng, 2000. Dynamic downscaling:
further results. RegClim General Technical Report, No.4. Available
from the Norwegian Institute for Air Research.
Hellström, C., D. Chen, C. Achberger and J. Räisänen, 2001.
Comparison of climate change scenarios for Sweden based on statistical and dynamical downscaling of monthly precipitation. Climate
Research, 19:45-55.
Herber, A., L.Thomason, H. Gernandt, U. Leiterer, D. Nagel, K.-H.
Schulz, J. Kaptur,T. Albrecht and J. Notholt, 2002. Continuous day
and night aerosol optical depth observations in the Arctic between
1991 and 1999. Journal of Geophysical Research, 107(D10):4097,
doi: 10.1029/2001JD000536.
Hewitson, B.C. and R.G. Crane, 1996. Climate downscaling: techniques
and application. Climate Research, 7:85-95.
Heyden, H., E. Zorita and H.Von Storch, 1996. Statistical downscaling
of monthly mean North Atlantic air-pressure to sea level anomalies in
the Baltic Sea.Tellus, 48A:312-323.
Heyden, H., H. Fock and W. Greve, 1998. Detecting relationships
between interannual variability in ecological time series and climate
using a multivariate statistical approach – a case study on Helgoland
Road zooplankton. Climate Research, 10:179-191.
Hirst, A., S.P. O’Farrell and H.B. Gordon, 2000. Comparison of a coupled ocean-atmosphere model with and without oceanic eddyinduced advection. Part I: Ocean spinup and control integrations.
Journal of Climate, 13:139-163.
Hoerling, M.P., J.W. Hurrell and T. Xu, 2001.Tropical origin for recent
North Atlantic climate change. Science, 292:90-92.
Hostetler, S.W., F. Giorgi, G.T. Bates and P.J. Bartlein, 1994. Lakeatmosphere feedbacks associated with paleolakes Bonneville and
Lahontan. Science, 263:665-668.
Hunke, E.C. and J.K. Dukowicz, 1997. An elastic-viscous-plastic model for
sea ice dynamics. Journal of Physical Oceanography, 27:1849-1867.
Huntingford, C., R.G. Jones, C. Prudhomme, R. Lamb and J.H.C.
Gash, 2003. Regional climate model predictions of extreme rainfall
for a changing climate. Quarterly Journal of the Royal
Meteorological Society, 129:1607-1621.
Huth, R., 1999. Statistical downscaling in central Europe: Evaluation of
methods and potential predictors. Climate Research, 13:91-101.
Huth, R. and J. Kysely, 2000. Constructing site-specific climate change
scenarios on a monthly scale using statistical downscaling.Theoretical
and Applied Climatology, 66:13-27.
IGPO, 2000. GEWEX Cloud System Study (GCSS) Second Science and
Implementation Plan. Global Energy and Water Cycle Experiment
(GEWEX). International GEWEX Project Office, Document No. 34,
52pp.
IPCC, 1994. IPCC Technical Guidelines for Assessing Climate Change
Impacts and Adaptations. Prepared by Working Group II.T.R. Carter,
M.L. Parry, H. Harasawa and S. Nishioka (eds.).WMO/UNEP,
CGER-IO15-94. Intergovernmental Panel on Climate Change, 59pp.
IPCC, 1996. Climate Change 1995:The Science of Climate Change.
Contribution of Working Group I to the Second Assessment Report
of the Intergovernmental Panel on Climate Change. J.T. Houghton,
L.G. Meira Filho, B.A. Callander, N. Harris, A. Kattenberg and K.
Maskell (eds.). Intergovernmental Panel on Climate Change.
Cambridge University Press, 572pp.
IPCC, 2001. Climate Change 2001:The Scientific Basis. Contribution of
Working Group I to the Third Assessment Report of the
Intergovernmental Panel on Climate Change. J.T. Houghton,Y. Ding,
D.J. Griggs, M. Noguer, P.J. van der Linden, X. Dai, K. Maskell and
C.A. Johnson (eds.). Intergovernmental Panel on Climate Change.
Cambridge University Press, 881pp.
IPCC-TGCIA, 1999. Guidelines on the Use of Scenario Data for
Climate Impact and Adaptation Assessment.Version 1. Prepared by
T.R. Carter, M. Hulme and M. Lal. Intergovernmental Panel on
Climate Change,Task Group on Scenarios for Climate and Impact
Assessment, 69pp.
Jacob, D. and R. Podzun, 1997. Sensitivity studies with the Regional
Climate Model REMO. Meteorology and Atmospheric Physics,
63:119-129.
Johns,T.C., 1996. A description of the second Hadley Centre coupled
model (HADCM2). Report No. 71. Hadley Centre for Climate
Prediction and Research, UK.
Johns,T.C., R.E. Carnell, J.F. Crossley, J.M. Gregory, J.F.B. Mitchell,
C.A. Senior, S.F.B.Tett and R.A.Wood, 1997.The second Hadley
Centre coupled atmosphere-ocean GCM: model description, spinup
and validation. Climate Dynamics, 13:103-134.
Jones, R.G., J.M. Murphy and M. Noguer, 1995. Simulation of climate
change over Europe using a nested regional climate model. I:
Assessment of control climate, including sensitivity to location of lateral boundaries. Quarterly Journal of the Royal Meteorological
Society, 121:1413-1449.
Jones, R.G., J.M. Murphy, M. Noguer and A.B. Keen, 1997. Simulation
of climate change over Europe using a nested regional-climate
model. II: comparison of driving and regional model responses to a
doubling of carbon dioxide. Quarterly Journal of the Royal
Meteorological Society, 123:265-292.
Kaas, E. and P. Frich, 1995. Diurnal temperature range and cloud cover
in the Nordic countries: observed trends and estimates for the
future. Atmospheric Research, 37:211-228.
Kalnay, E., 2003. Atmospheric Modeling, Data Assimilation and
Predictability. Cambridge University Press, 341pp.
Kattsov,V.M. and V.P. Meleshko, 2004. Evaluation of atmosphere-ocean
general circulation models used for projecting future climate change.
Izvestia, Russian Academy of Sciences – Atmospheric and Oceanic
Physics, 40(6):647-658.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Kattsov,V.M. and J.E.Walsh, 2002. Reply to comments on ‘Twentiethcentury trends of Arctic precipitation from observational data and a
climate model simulation’ by H. Paeth, A. Hense, and R.
Hagenbrock. Journal of Climate, 15:804-805.
Kattsov,V.M.,V.P. Meleshko,V.M. Gavrilina,V.A. Govorkova and T.V.
Pavlova, 1998. Freshwater budget of the polar regions as simulated
with current atmospheric general circulation models. Izvestia Russian
Academy of Sciences Phys. Atmos. Ocean, 34:479-489.
Kattsov,V.M., J.E.Walsh, A. Rinke and K. Dethloff, 2000. Atmospheric
climate models: simulations of the Arctic Ocean fresh water budget
components. In: E.L. Lewis, E.P. Jones, P. Lemke,T.D. Prowse and P.
Wadhams (eds.).The Freshwater Budget of the Arctic Ocean, pp.
209-247. Kluwer Academic Publishers.
Kattsov,V.M., S.V.Vavulin,V.A. Govorkova and T.V. Pavlova, 2003.
Scenarios of the Arctic climate change in the 21st century. Russian
Meteorology and Hydrology, 10:5-19.
Katz, R.W. and M.B. Parlange, 1996. Mixtures of stochastic processes:
applications to statistical downscaling. Climate Research, 7:185-193.
Källén, E.,V. Kattsov, J.Walsh and E.Weatherhead, 2001. Report from
the Arctic Climate Impact Assessment Modeling and Scenarios
Workshop. Stockholm, January 29-31 2001. 35pp.
Khrol,V.P. (ed.), 1996. Atlas of Water Balance of the Northern Polar
Area. Gidrometeoizdat, St. Petersburg, 81pp.
Kidson, J.W. and C.S.Thompson, 1998. Comparison of statistical and
model-based downscaling techniques for estimating local climate
variations. Journal of Climate, 11:735-753.
Kiilsholm, S., J.H. Christensen, K. Dethloff and A. Rinke, 2003. Net
accumulation of the Greenland Ice Sheet: Modelling Arctic regional
climate change. Geophysical Research Letters, 30,
10.1029/2002GL015742.
Kistler, R., E. Kalnay,W. Collins, S. Saha, G.White, J.Woollen, M.
Chelliah,W. Ebisuzaki, M. Kanamitsu,V. Kousky, H. van den Dool,
R. Jenne and M. Fiorino, 2001.The NCEP-NCAR 50-year reanalysis: Monthly means CD-ROM and documentation. Bulletin of the
American Meteorological Society, 82:247-268.
Knippetz, P., U. Ulbrich and P. Speth, 2000. Changing cyclones and surface wind speeds over the North Atlantic and Europe in a transient
GHG experiment. Climate Research, 15:109-122.
Knutson,T.R.,T.L. Delworth, K.W. Dixon and R.J. Stouffer, 1999.
Model assessment of regional surface temperature trends (19491997). Journal of Geophysical Research, 104:30981-30996.
Langley Atmospheric Sciences Data Center, 1983-1991. Langley EightYear Shortwave and Longwave Surface Radiation Budget Dataset,
July 1983-June 1991. NASA. (CD-ROM)
Laprise, R., D. Caya, M. Giguère, G. Bergeron, H. Côte, J.-P. Blanchet,
G.J. Boer and N.A. McFarlane, 1998. Climate and climate change in
western Canada as simulated by the Canadian regional climate
model. Atmosphere-Ocean, 36:119-167.
Legates, D.R. and C.L.Willmott, 1990. Mean seasonal and spatial variability in gauge-corrected global precipitation. International Journal
of Climatology, 10:111-133.
Leggett, J.W., J. Pepper and R.J. Swart, 1992. Emission scenarios for the
IPCC: an update. In: J.T. Houghton, B.A. Callander and S.K.Varney
(eds.). Climate Change 1992.The Supplementary Report to the IPCC
Scientific Assessment, pp. 69-95. Cambridge University Press.
Lemke, P.,W.D. Hibler III, G.M. Flato, M. Harder and M. Kreyscher,
1997. On the improvement of sea-ice models for climate simulations:The sea-ice model intercomparison project. Annals of
Glaciology, 25:183-187.
Linderson, M-L., C. Achberger and D. Chen, 2004. Statistical downscaling and scenario construction of precipitation in Scania, southern
Sweden. Nordic Hydrology, 35(3):261-278.
Lorenz, E.N., 1963.The predictability of hydrodynamic flow.Transactions
of the NewYork Academy of Sciences, Series II, 25:409-432.
Lunkeit, F., M. Ponater, R. Sausen, M. Sogalla, U. Ulbrich and M.
Windelband, 1996. Cyclonic activity in a warmer climate. Beitrage
zur Physik der Atmosphare, 69:393-407.
Lynch, A.H.,W.L. Chapman, J.E.Walsh and G.Weller, 1995.
Development of a regional climate model of the western Arctic.
Journal of Climate, 8:1555-1570.
Lynch, A.H., M.F. Glück,W.L. Chapman, D.A. Bailey and J.E.Walsh,
1997. Remote sensing and climate modeling of the St. Lawrence Is.
Polynya.Tellus, 49A:277-297.
Lynch, A.H., D.L. McGinnes and D.A. Bailey, 1998. Snow-albedo and the
spring transition in a regional climate system model: influence of land
surface model. Journal of Geophysical Research, 103:29037-29049.
Lynch, A.H., J.A. Maslanik and W.Wu, 2001. Mechanisms in the development of anomalous sea ice extent in the western Arctic: A case
study. Journal of Geophysical Research, 106:28097-28105.
Mabuchi, K.,Y. Sato and H. Kida, 2000. Numerical study of the relationships between climate and the carbon dioxide cycle on a regional
scale. Journal of the Meteorological Society of Japan, 78:25-46.
147
Machenhauer, B., M.Windelband, M. Botzet, R. Jones and M. Deque,
1996.Validation of present-day regional climate simulations over
Europe: Nested LAM and variable resolution global model simulations with observed or mixed layer ocean boundary conditions. MPI
Report 191. Max Planck Institute for meteorology, Hamburg.
Machenhauer, B., J.Windelband, M. Botzet, J.H. Christensen, M.
Deque, R. Jones, P.M. Ruti and G.Visconti, 1998.Validation and
analysis of regional present-day climate and climate change simulations over Europe. MPI Report 275. Max Planck Institute for meteorology, Hamburg.
Majewski, D., D. Liermann, P. Prohl, B. Ritter, M. Buchhold,T. Hanisch,
G. Paul,W.Wergen and J. Baumgardner, 2002.The operational global
Icosahedral-Hexagonal Gridpoint Model GME: description and highresolution tests. Monthly Weather Review, 130:319-338.
Mahrt, L., 1998: Stratified atmospheric boundary layers and breakdown of
models.Theoretical and Computational Fluid Dynamics, 11:263-279.
Manabe, S. and R.J. Stouffer, 1997. Coupled ocean-atmosphere model
response to freshwater input: comparision to Younger Dryas event.
Paleoceanography, 12:321-336.
Manabe, S. and R.T.Wetherald, 1975.The effects of doubling the CO2
concentration on the climate of a general circulation model. Journal
of Atmospheric Sciences, 32:3-15.
Manabe, S., R.J. Stouffer, M.J. Spelman and K. Bryan, 1991.Transient
responses of a coupled ocean-atmosphere model to gradual changes
of atmospheric CO2. Part I: Annual mean response. Journal of
Climate, 4:785-818.
Martin, E., B.Timbal and E. Brun, 1997. Downscaling of general circulation model output: simulation of snow climatology of the French
Alps and sensitivity to climate change. Climate Dynamics, 13:45-56.
Maslanik, J.A., M.C. Serreze and R.G. Barry, 1996: Recent decreases in
Arctic summer ice cover and linkage to atmospheric circulation
anomalies. Geophysical research Letters, 23:1677-1680.
Maslanik, J.A., A.H. Lynch, M.C. Serreze and W.Wu, 2000. A case
study of regional climate anomalies in the Arctic: performance
requirements for a coupled model. Journal of Climate, 13:383-401.
McAvaney, B.J., C. Covey, S. Joussaume,V. Kattsov, A. Kitoh,W. Ogana,
A.J. Pitman, A.J.Weaver, R.A.Wood and Z.-C. Zhao, 2001. Model
evaluation. In: J.T. Houghton,Y. Ding, D.J. Griggs, M. Noguer, P.J. van
der Linden, X. Dai, K. Maskell and C.A. Johnson (eds.). pp. 471-524.
Climate Change 2001:The Scientific Basis. Contribution of Working
Group I to the Third Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press.
McGregor, J.L., J.J. Katzfey and K.C. Nguyen, 1999. Recent regional
climate modelling experiments at CSIRO. In: H. Ritchie (ed.).
Research Activities in Atmospheric and Oceanic Modelling, pp. 7.377.38. CAS/JSC Working Group on Numerical Experimentation
Report 28.WMO/TD - No. 942.World Meteorological
Organization, Geneva.
Mearns, L.O., M. Hulme,T.R. Carter, R. Leemans, M. Lal and P.
Whetton, 2001. Climate scenario development. In: J.T. Houghton,Y.
Ding, D.J. Griggs, M. Noguer, P.J. van der Linden, X. Dai, K.
Maskell and C.A. Johnson (eds.). pp. 739-768. Climate Change
2001:The Scientific Basis. Contribution of Working Group I to the
Third Assessment Report of the Intergovernmental Panel on Climate
Change. Cambridge University Press.
Meehl, G.A., G.J. Boer, C. Covey, M. Latif and R.J. Stouffer, 2000.The
Coupled Model Intercomparison Project (CMIP). Bulletin of the
American Meteorological Society, 81:313-318.
Meleshko,V.P.,V.M. Kattsov,V.A. Govorkova, S.P. Malevsky-Malevich,
E.D. Nadyozhina and P.V. Sporyshev, 2004. Anthropogenic climate
changes in the 21st century in Northern Eurasia. Russian
Meteorology and Hydrology, 7:5-26.
Murphy, J., 1999. An evaluation of statistical and dynamical techniques
for downscaling local climate. Journal of Climate, 12:2256-2284.
Murphy, J., 2000. Predictions of climate change over Europe using statistical and dynamical downscaling techniques. International Journal
of Climatology, 20:489-501.
Murray, R.J., 1996. Explicit generation of orthogonal grids for ocean
models. Journal of Computational Physics, 126:251-273.
Naki5enovi5, N. and R. Swart (eds.), 2000. Intergovernmental Panel on
Climate Change, Special Report on Emissions Scenarios. Cambridge
University Press, 599pp.
NAST, 2001. Climate Change Impacts on the United States:The
Potential Consequences of Climate Variability and Change. National
Assessment Synthesis Team, Report for the US Global Change
Research Program. Cambridge University Press, 620pp.
Nelson, F.E. and S.I. Outcalt, 1987. A computational method for prediction and regionalization of permafrost. Arctic and Alpine Research,
19:279-288.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
148
Arctic Climate Impact Assessment
New, M., M. Hulme and P. Jones, 1999. Representing twentieth-century
space-time climate variability. Part I: Development of a 1961-90 mean
monthly terrestrial climatology. Journal of Climate, 12:829-856.
New, M., M. Hulme and P. Jones, 2000. Representing twentieth-century
space-time climate variability. Part II: Development of 1901-96
monthly grids of terrestrial surface climate. Journal of Climate,
13:2217-2238.
Noguer, M., R. Jones and J. Murphy, 1998. Sources of systematic errors
in the climatology of a regional climate model over Europe. Climate
Dynamics, 14:691-712.
Nozawa,T., S. Emori,T.Takemura,T. Nakajima, A. Numaguti, A. AbeOuchi and M. Kimoto, 2000. Coupled ocean-atmosphere model
experiments of future climate change based on IPCC SRES scenarios.
Preprints, Eleventh Symposium on Global Change Studies, 9-14
January 2000, Long Beach, California, pp. 352-355.
Omstedt, A. and D. Chen, 2001. Influence of atmospheric circulation on
the maximum ice extent in the Baltic Sea. Journal of Geophysical
Research, 106(C3):4493-4500.
Otterå, O.H., H. Drange, M. Bentsen, N.G. Kvamstø and D. Jiang,
2003:The sensitivity of the present day Atlantic meridional overturning circulation to freshwater forcing. Geophysical Research Letters,
30: 1898, doi:101029/2003GL017578.
Pacanowski, R.C. and S.M. Griffies, 1999.The MOM 3 Manual.
Geophysical Fluid Dynamics Laboratory/NOAA, Princeton, New
Jersey, 680pp.
Palmer,T.N. and J. Räisänen, 2002. Quantifying the risk of extreme seasonal precipitation events in a changing climate. Nature, 415:512-514.
Palutikof, J.P., C.M. Goodess, S.J.Watkins and T. Holt, 2002. Generating
daily rainfall and temperature scenarios at multiple sites: Examples
from the Mediterranean. Journal of Climate, 15:3529-3548.
Paterson, A.B. and N. Reeh, 2001.Thinning of the ice sheet in northwest Greenland over the past forty years. Nature, 414:60-62.
Peterson, B.J., R.M. Holmes, J.W. McClelland, C.J.Vörösmarty, R.B.
Lammers, A.I. Shiklomanov, I.A. Shiklomanov and S. Rahmstorf,
2002. Increasing river discharge to the Arctic Ocean. Science,
298:2171-2173.
Polyakov, I. and M. Johnson, 2000. Arctic decadal and interdecadal variability. Geophysical Research Letters, 24:4097-4100.
Polyakov, I., S.-I. Akasofu, U. Bhatt, R. Colony, M. Ikeda, A. Makshtas,
C. Swingley, D.Walsh and J.Walsh, 2002a.Trends and variations in
Arctic climate system. Eos,Transactions, American Geophysical
Union, 83(47):547-548.
Polyakov, I., G. Alekseev, R. Bekryaev, U. Bhatt, R. Colony, M. Johnson,
V. Karklin, A. Makshtas, D.Walsh and A.Yulin, 2002b.
Observationally based assessment of polar amplification of global
warming. Geophysical Research Letters, 29(18):1878-1881.
Poulus, G.S., and S.P. Burns, 2003. An evaluation of bulk Ri-based surface layer flux formulas for stable and very stable conditions with
intermittent turbulence. Journal of Atmospheric Science, 60:25232537.
Power, S.B., R.A. Colman, B.J. McAvaney, R.R. Dahni, A.M. Moore and
N.R. Smith, 1993.The BMRC coupled atmosphere/ocean/sea-ice
model. Bureau of Meteorology Research Centre, Research Rep. 37.
Melbourne, 58pp.
Proshutinsky, A., M. Steele, J. Zhang, G. Holloway, N. Steiner, S.
Hakkinen, D. Holland, R. Gerdes, C. Koeberle, M. Karcher, M.
Johnson,W. Maslowski,W.Walczowski,W. Hibler and J.Wang, 2001.
Multinational effort studies differences among Arctic Ocean models.
Eos,Transactions, American Geophysical Union, 82(637):643-644.
Puri, K., 2002. Activities of the CAS/JSC Working Group on Numerical
Experimentation (WGNE). In: H. Ritchie (ed.). Research Activities
in Atmospheric and Oceanic Modelling. Report No. 32,WMO/TDNo. 1105.
Qian,Y. and F. Giorgi, 1999. Interactive coupling of regional climate and
sulfate aerosol models over East Asia. Journal of Geophysical
Research, 104:6501-6514.
Quinn, P.K.,T.L. Miller,T.S. Bates, J.A. Ogren, E. Andrews and G.E.
Shaw, 2002. A 3-year record of simultaneously measured aerosol
chemical and optical properties at Barrow, Alaska. Journal of
Geophysical Research, 107(D11):4130,
doi:10.1029/2001JD001248.
Randall, D., J. Curry, D. Battisti, G. Flato, R. Grumbine, S. Hakkinen,
D. Martinson, R. Preller, J.Walsh and J.Weatherly, 1998. Status of
and outlook for large-scale modeling of atmosphere-ice-ocean interactions in the Arctic. Bulletin of the American Meteorological
Society, 79:197-219.
Räisänen, J., 2001. CO2-induced climate change in the Arctic area in the
CMIP2 experiments. SWECLIM Newsletter, 11:23-28.
Räisänen, J. and R. Döscher, 1999. Simulation of present-day climate in
Northern Europe in the HadCM2 GCM. Reports on Meteorology and
Climatology, No. 48. Swedish Meteorological and Hydrological Institute.
Räisänen, J. and T.N. Palmer, 2001. A probability and decision-model
analysis of a multi-model ensemble of climate change simulations.
Journal of Climate, 14:3212-3226.
Räisänen, J., M. Rummukainen, A. Ullerstig, B. Bringfelt, U. Hansson
and U.Willén, 1999.The first Rossby Centre regional climate scenario – dynamical downscaling of CO2-induced climate change in the
HadCM2 GCM. Reports on Meteorology and Climatology, No. 85,
Swedish Meteorological and Hydrological Institute, 56pp.
Räisänen, J., U. Hansson, A. Ullerstig, R. Döscher, L.P. Graham, C.
Jones, M. Meier, P. Samuelsson and U.Willén, 2003. GCM driven
simulations of recent and future climate with the Rossby Centre coupled atmosphere Baltic Sea regional climate model RCAO. Reports
on Meteorology and Climatology, No. 101, Swedish Meteorological
and Hydrological Institute, 61pp.
Rind, D., P. deMenocal, G. Russell, S. Sheth, D. Collins, G. Schmidt and
J.Teller, 2001. Effects of glacial meltwater in the GISS coupled atmosphere-ocean model 1. North Atlantic Deep Water response. Journal
of Geophysical Research, 16:27335-27353.
Rinke, A., A.H. Lynch and K. Dethloff, 2000. Intercomparison of Arctic
regional climate simulations: case studies of January and June 1990.
Journal of Geophysical Research, 105:29669-29683.
Rinke, A., R. Gerdes, K. Dethloff,T. Kandlbinder, M. Karcher, F.
Kauker, S. Frickenhaus, C. Köberle and W. Hiller, 2003. A case study
of the anomalous Arctic sea ice conditions during 1990: insight from
coupled and uncoupled regional climate model simulations. Journal
of Geophysical Research, 108 (D9),4275,
doi:10.1029/2002JD003146.
Roeckner, E., J.M. Oberhuber, A. Bacher, M. Christoph and I. Kirchner,
1996. ENSO variability and atmospheric response in a global coupled
atmosphere-ocean GCM. Climate Dynamics, 12:737-754.
Roeckner, E., L. Bengtsson, J. Feichter, J. Lelieveld and H. Rodhe,
1999.Transient climate change simulations with a coupled atmosphere-ocean GCM including the tropospheric sulfur cycle. Journal
of Climate, 12:3004-3032.
Roed, L.P., X.B. Shi and B. Hackett, 2000.The Importance of Allowing
Turbulent and Diffusive Diapycnal Mixing in Isopycnic Coordinate
Ocean Models. RegClim General Tech. Rep. 4, 139-148. Available
from the Norwegian Institute for Air Research.
Rummukainen, M., J. Räisänen, B. Bringfelt, A. Ullerstig, A. Omstedt, U.
Willén, U. Hansson and C. Jones, 2001. A regional climate model for
northern Europe: model description and results from the downscaling
of two GCM control simulations. Climate Dynamics, 17:339-359.
Ruosteenoja, K.,T.R. Carter, K. Jylhä and H.Tuomenvirta, 2003.
Future climate in world regions: an intercomparison of model-based
projections for the new IPCC emissions scenarios.The Finnish
Environment 644, Finnish Environment Institute.
Russell, G.L. and D. Rind, 1999. Response to CO2 transient increase in
the GISS coupled model: regional coolings in a warmer climate.
Journal of Climate, 12:531-539.
Rysgaard, S.,T.Vang, M. Stjernholm, B. Rasmussen, A.Windelin and S.
Kiilsholm, 2003. Physical conditions, carbon transport and climate
change impacts in a NE Greenland fjord. Arctic, Antarctic and Alpine
Research, 35(3):301-312.
Sadourny, R., A. Arakawa and Y. Mintz, 1968. Integration of the nondivergent barotropic equation with an icosahedral hexagonal grid on
the sphere. Monthly Weather Review, 96:351-356.
Sailor, D.J. and X. Li, 1999. A semiempirical downscaling approach for
predicting regional temperature impacts associated with climatic
change. Journal of Climate, 12:103-114.
Santer, B.D.,T.M.L.Wigley, M.E. Schlesinger and J.F.B. Mitchell, 1990.
Developing climate scenarios from equilibrium GCM results. MPI
Rep. 47, Max Planck Institute for Meteorology, Hamburg, 29pp.
Sausen, R., K. Barthel and K. Hasselmann, 1987. A flux correction
method for removing the climate drift of climate models. Report
No. 1. Max Planck Institute for Meteorology, 39pp.
Schiller, A., U. Mikolajewicz and R.Voss, 1997.The stability of the
North Atlantic thermohaline circulation in a coupled ocean-atmosphere general circulation model. Climate Dynamics, 13:325-347.
Schnur, R. and D. Lettenmaier, 1997. A case study of statistical downscaling in Australia using weather classification by recursive partitioning. Journal of Hydrology, 211:362-379.
Schoof, J.T. and S.C. Pryor, 2001. Downscaling temperature and precipitation: A comparison of regression-based methods and artificial neural networks. International Journal of Climatology, 21:773-790.
Schrum, C., U. Huebner, D. Jacob and R. Podzun, 2001. A coupled
atmosphere/ice/ocean model for the North Sea and the Baltic Sea.
Berichte aus dem Zentrum für Meeres- und Klimaforschung Nr. 41,
Reihe B: Ozeanographie. Inst. Meereskunde, Hamburg.
Schubert, S., 1998. Downscaling local extreme temperature changes in
south-eastern Australia from the CSIRO MARK2 GCM. International
Journal of Climatology, 18:1419-1439.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
Chapter 4 • Future Climate Change:Modeling and Scenarios for the Arctic
Schweiger, A.J., R.W. Lindsay, J.R. Key and J.A. Francis, 1999. Arctic
clouds in multiyear satellite data sets. Geophysical Research Letters,
26:1845-1848.
Semenov, M.A. and E. Barrow, 1997. Use of stochastic weather generator in the development of climate change scenarios. Climatic
Change, 35:397-414.
Semenov, M.A., R.J. Brooks, E.M. Barrow and C.W. Richardson, 1998.
Comparison of the WGEN and LARS-WG stochastic weather generators for diverse climates. Climate Research, 10:95-107.
Shindell, D.T., G.A. Schmidt, R.L. Miller and D. Rind, 2001. Northern
Hemisphere winter climate response to greenhouse gas, ozone, solar,
and volcanic forcing. Journal of Geophysical Research, 106:7193-7210.
Small, E.E., F. Giorgi and L.C. Sloan, 1999a. Regional climate model
simulation of precipitation in central Asia: Mean and interannual variability. Journal of Geophysical Research, 104:6563-6582.
Small, E.E., L.C. Sloan, S. Hostetler and F. Giorgi, 1999b. Simulating
the water balance of the Aral Sea with a coupled regional climatelake model. Journal of Geophysical Research, 104:6583-6602.
Smith, J.B., M. Hulme, J. Jaagus, S. Keevallik, A. Mekonnen and K.
Hailemariam, 1998. Climate change scenarios. In: J.F. Feenstra, I.
Burton, J. Smith and R.S.J.Tol (eds.). Handbook on Methods for
Climate Change Impact Assessment and Adaptation Strategies, pp. 31 - 3-40.Version 2.0. United Nations Environment Programme and
Institute for Environmental Studies,Vrije Universiteit, Amsterdam.
Solman, S.A. and M.N. Nuñez, 1999. Local estimates of global climate
change: A statistical downscaling approach. International Journal of
Climatology, 19:835-861.
Stendel, M.,T. Schmith, E. Roeckner and U. Cubasch, 2000.The climate of the 21st century: transient simulations with a coupled atmosphere-ocean general circulation model. Danish Climate Centre
Report 00-6, 51pp.
Stocker,T.F., G.K.C. Clarke, H. Le Treut, R.S. Lindzen,V.P. Meleshko,
R.K. Mugara,T.N. Palmer, R.T. Pierrehumbert, P.J. Sellers, K.E.
Trenberth and J.Willebrand, 2001. Physical climate processes and
feedbacks. In: J.T. Houghton,Y. Ding, D.J. Griggs, M. Noguer, P.J. van
der Linden, X. Dai, K. Maskell and C.A. Johnson (eds.), pp. 418-470.
Climate Change 2001:The Scientific Basis. Contribution of Working
Group I to the Third Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press.
Stott, P., 2003: Attribution of regional-scale temperature changes to
anthropogenic and natural causes. Geophysical Research Letters,
30(14): 1728, doi:10.1029/2003GL017324.
Stott, P.A., S.F.B.Tett, G.S. Jones, M.R. Allen, J.F.B. Mitchell and G.J.
Jenkins, 2000. External control of twentieth century temperature variations by natural and anthropogenic forcings. Science, 15:2133-2137.
Stouffer, R.J. and K.W. Dixon, 1998. Initialization of coupled models
for use in climate studies. In: A. Staniforth (ed.). Research Activities
in Atmospheric and Oceanic Modelling, pp. I.1-I.15. Rep. 27,
WMO/TD-No. 865.
Tabazadeh, A., E.J. Jensen, O.B.Toon, K. Drdla and M.R. Schoeberl,
2001. Role of the stratospheric polar freezing belt in denitrification.
Science, 291:2591-2594.
Takahashi, M., 1999. Simulation of the quasibiannial oscillation in a general circulation model. Geophysical Research Letters, 26, 1307-1310.
Takle, E.S.,W.J. Gutowski, R.W. Arritt, Z. Pan, C.J. Anderson, R.S. da
Silva, D. Caya, S.-C. Chen, F. Giorgi, J.H. Christensen, S.-Y. Hong,
H.-M.H. Juang, J. Katzfey,W.M. Lapenta, R. Laprise, P. Lopez, G.E.
Liston, J. McGregor, A. Pielke and J.O. Roads, 1999. Project to
intercompare regional climate simulation (PIRCS): Description and
initial results. Journal of Geophysical Research, 104:19443-19461.
Tao, X., J.E.Walsh and W.L. Chapman, 1996. An assessment of global
climate model simulations of Arctic air temperatures. Journal of
Climate, 9:1060-1076.
Tett, S.F.B., G.S. Jones, P.A. Stott, D.C. Hill, J.F.B. Mitchell, M.A.Allen,
W.J. Ingram,T.C. Johns, C.E. Johnson,A. Jones, D.L. Roberts, D.M.H.
Sexton and M.J.Woodage, 2000. Estimation of Natural and
Anthropogenic Contributions to 20th Century. Hadley Centre Tech Note
19. Hadley Centre for Climate Prediction and Response, UK, 52pp.
Tokioka,T., A. Noda, A. Kitoh,Y. Nikaidou, S. Nakagawa,T. Motoi, S.
Yukimoto and K.Takata, 1995. A transient CO2 experiment with the
MRI CGCM. Quick Report. Journal of the Meteorological Society of
Japan, 73(4):817-826.
Trenberth, K. (ed.), 1992. Climate System Modelling. Cambridge
University Press, 788pp.
Tsvetsinskaya, E., L.O. Mearns and W.E. Easterling, 2000. Effects of
plant growth and development on interannual variability in
mesoscale atmospheric simulations. In: Proceedings of the Tenth
International Offshore and Polar Engineering Conference.Vol. I, pp.
729-736. International Society of Offshore and Polar Engineers,
Cupertino, California.
149
Ulbrich, U. and M. Christoph, 1999. A shift of the NAO and increasing
storm track activity over Europe due to anthropogenic greenhouse
gas forcing. Climate Dynamics, 15:551-559.
Van den Brink, H.W., G.P. Konnen and J.D. Opsteegh, 2003.The reliability of extreme surge levels, estimated from observational records
of order of hundred years. Journal of Coastal Research, 19:376-388.
Vellinga, M., R.A.Wood and J.M. Gregory, 2002. Processes governing
the recovery of a perturbed thermohaline circulation in HadCM3.
Journal of Climate, 15:764-780.
von Storch, H., 1999.The global and regional climate system. In: H.
von Storch and G. Flöser (eds.). Anthropogenic Climate Change, pp.
3-36. Springer Verlag.
Voss, R., R. Sausen and U. Cubasch, 1998. Periodically synchronously
coupled integrations with the atmosphere-ocean general circulation
model ECHAM3/LSG. Climate Dynamics, 14:249-266.
Wallace, J.M., 2000. North Atlantic Oscillation/annular mode:Two paradigms – one phenomenon. Quarterly Journal of the Royal
Meteorological Society, 126:729-805.
Walsh, J.E., in press. Summary of a workshop on modeling the Arctic
atmosphere, Madison,Wisconsin, 20-22 May 2002. Bulletin of the
American Meteorological Society.
Walsh, J.E.,V. Kattsov, D. Portis and V. Meleshko, 1998. Arctic precipitation and evaporation: model results and observational estimates.
Journal of Climate, 11:72-87.
Walsh, J.E.,V. Kattsov,W. Chapman,V. Govorkova and T. Pavlova, 2002.
Comparison of Arctic climate simulations by uncoupled and coupled
global models. Journal of Climate, 15:1429-1446.
Washington,W.M., J.W.Weatherly, G.A. Meehl, A.J. Semtner Jr.,T.W.
Bettge, A.P. Graig,W.G. Strand Jr., J. Arblaster,V.B.Wayland, R.
James and Y. Zhang, 2000. Parallel climate model (PCM) control and
transient simulations. Climate Dynamics, 16:755-774.
WCRP, 2002.WOCE/CLIVAR, 2002:WOCE/CLIVAR Working Group
on Ocean Model Development. Report of the Third Session.World
Climate Research Programme, Informal Rep. No. 14/2002.
Weisse, R., H. Heyen and H. von Storch, 2000. Sensitivity of a regional
atmospheric model to a sea state-dependent roughness and the need
for ensemble calculations. Monthly Weather Review, 128:3631-3642.
Wilby, R.L. and T.M.L.Wigley, 1997. Downscaling general circulation
model output: a review of methods and limitations. Progress in
Physical Geography, 21:530-548.
Wilby, R.L. and T.M.L.Wigley, 2000. Precipitation predictors for downscaling: observed and general circulation model relationships.
International Journal of Climatology, 20:641-661.
Wilby, R.L.,T.M.L.Wigley, D. Conway, P.D. Jones, B.C. Hewitson, J.
Main and D.S.Wilks, 1998. Statistical downscaling of general circulation model output: A comparison of methods.Water Resources
Research, 34:2995-3008.
Wilby, R.L., C.W. Dawson and E.M. Barrow, 2002. SDSM – a decision
support tool for the assessment of regional climate change impacts.
Environmental Modelling and Software, 17:145-157.
Wilks, D., 1999. Multisite downscaling of daily precipitation with a stochastic weather generator. Climate Research, 11:125-136.
Yukimoto, S., M. Endoh,Y. Kitamura, A. Kitoh,T. Motoi and A. Noda,
2000. ENSO-like interdecadal variability in the Pacific Ocean as simulated in a coupled GCM. Journal of Geophysical Research,
105(D10):13945-13963.
Pre-release Version 24 March, 2005
www.acia.uaf.edu
150
Arctic Climate Impact Assessment
Pre-release Version 24 March, 2005
www.acia.uaf.edu